design and test of a highly scalable servo drive system...
TRANSCRIPT
Fachbereich 18
Elektrotechnik und Informationstechnik
Design and test of a highly scalable servo drive system based on PM linear synchronous motors
genehmigte
Dissertation von
Dipl.-Ing. Sorin Mihail Silaghiu aus Ploiesti/Rumänien
Referent: Prof. Dr.-Ing. Peter Mutschler
Korreferent: Prof. Dr.-Ing. Bernhard Piepenbreier
Tag der Einreichung: 27.10.2011
Tag der mündlichen Prüfung: 15.03.2012
D17
Darmstadt 2012
Design and test of a highly scalable
servo drive system based on PM linear synchronous motors
Vom Fachbereich 18
Elektrotechnik und Informationstechnik
der Technischen Universität Darmstadt
zur Erlangung des akademischen Grades eines
Doktor-Ingenieurs (Dr.-Ing.)
genehmigte
Dissertation
von
Dipl.-Ing. Sorin Mihail Silaghiu
geboren am 31. August 1983 in Ploiesti, Rumänien
Referent: Prof. Dr.-Ing. Peter Mutschler
Korreferent: Prof. Dr.-Ing. Bernhard Piepenbreier
Tag der Einreichung: 27. Oktober 2011
Tag der mündlichen Prüfung: 15. März 2012
D17
Darmstadt 2012
I
Preface
This PhD thesis is the result of my research work at the department of Power Electronics and
Control of Drives at Darmstadt University of Technology.
My special thanks go first to the head of the department, Prof. Dr.-Ing. Peter Mutschler, for
guiding this work. I am very grateful for his valuable advices, patience and encouragement during
all the research years.
I would like to express my sincere gratitude to Prof. Dr.-Ing. Bernhard Piepenbreier from the
department of Electrical Drives and Machines at the Friedrich-Alexander University in Erlangen-
Nürnberg for his interest and willingness to be the co-reviewer of this thesis.
For the financial support of my project, MU 1109/20-1, I would like to thank DFG (Deutsche
Forschungsgemeinschaft).
I thank my colleagues and also the students, who did their seminar or diploma thesis in my
research field at the department, for the professional discussions, suggestions and support. The
realisation of the experimental set-up would not have been possible without the support of the
technical staff. I also appreciate the help of our secretary regarding all administrative issues. I am
grateful to all my colleagues at the department for the very good working atmosphere.
I want to give my sincere thanks also to my parents, especially my mother, who supported and
encouraged me throughout my entire education.
At the end I thank God for giving me the opportunity to do my PhD in Darmstadt, for His guidance
and for giving me the daily energy and strength that I need.
Darmstadt, 27.10.2011
III
Abstract
In many industrial plants materials have to be transported between several processing stations,
where they have to be processed with a high level of accuracy and precision. In the last years,
linear electrical drives, especially the long-stator linear motors are used for these types of
applications for both processing and transportation tasks. They have higher dynamics and
processing precision and lower maintenance costs compared to conventional systems, which
require additional mechanical gears.
The linear drive system must be modular and highly scalable in order to cover a wide range of
applications. For this reason, the track of the plant is made of several stator segments. The
excitation part of the motor is represented by permanent magnets (passive vehicles). Each stator
segment has a dedicated inverter (Power Processing Unit) and processor (Information Processing
Unit). Cheap IPMs (Intelligent Power Modules) are nowadays a good solution for implementing the
inverter. A DSP was used as processor. The DSP and the IPM are the main components of the
designed servo-controller, which together with a stator segment represents a module of the
system. The DSP controls the inverter and is also used for the communication with the DSPs of the
adjacent modules.
By connecting the ground potential of all servo-controllers to the negative DC-link rail (≈ -280 V),
a significant reduction in the implementation costs of the servo-controller was achieved.
When a vehicle crosses from one stator segment to the adjacent one, the control tasks migrate
physically in that respective adjacent DSP. Data exchange is therefore required within each cycle
of the current control loop (100 μs) between the adjacent modules. For an arbitrary scalable
modular system, there will be also an arbitrary high communication demand. This demand can
only be solved by a direct (Point-to-Point) connection between the adjacent DSPs. This connection
was realised by means of the cost-effective RS485 data transmission protocol.
A central control unit is responsible then for the cyclical (1-10 ms) generation of new position
reference values for the vehicles, according to a predefined schedule. The monitoring (assessment
of internal variables of the distributed servo-controllers) of the entire system is realised also in
real-time. Off-line download and upload actions of firmware or general data is also possible. For
these tasks, the communication between the central unit (PC) and the distributed servo-
controllers was realised by means of the Ethernet-based fieldbus EtherCAT.
Inside the processing stations of the system, the positioning accuracy and precision as well as the
dynamic have to be very good. For this reason, inside those stations, position sensors must be
IV
used. Outside those stations, for material transportation only, an EMF-based sensorless control
was implemented. This will further reduce the overall system costs.
A small section of such modular and highly scalable system was realised as an experimental set-up
in the context of this work, in order to test the functionality and the reliability of the proposed
system.
V
Kurzfassung
In vielen Industrieanlagen werden Materialien in Bearbeitungsstationen mit hoher Präzision
bearbeitet und zwischen diesen Stationen transportiert. Dafür kommen in den letzten Jahren
immer häufiger Linearantriebe, speziell der Langstator Linearmotor, sowohl für die Bearbeitung
als auch für das Transportieren zum Einsatz. Der Grund hierfür liegt in der hohen Dynamik, bei
gleichzeitig hoher Positioniergenauigkeit und in der Reduzierung der Wartungskosten, weil keine
zusätzlichen Getriebe notwendig sind, wie bei den rotierenden Antrieben.
Das Linear-Antriebsystem muss modular und beliebig skalierbar sein, um eine große Vielfalt von
Anwendungen abzudecken. Dafür wird der Fahrweg in viele Statorabschnitte unterteilt. Die
Sekundärteile (passive Fahrzeuge) sind Permanentmagneten. Jedem Statorsegment werden
sowohl ein Umrichter (Leistungsteil) als auch ein eigener Prozessor (Informationsteil) zugeordnet.
Preisgünstige IPMs (Intelligent Power Modules) bieten heutzutage eine gute Lösung für die
Implementierung des Umrichters. Ein DSP wird als Prozessor eingesetzt. Der DSP und der IPM
sind die Hauptkomponenten des ausgelegten Servocontrollers, der zusammen mit einem
Statorsegment ein Modul des Systems darstellt. Der DSP dient sowohl zur Steuerung des
Umrichters als auch zur Kommunikation mit den benachbarten Modulen.
Eine deutliche Reduzierung der Implementierungskosten des Servocontrollers wurde
gewährleistet, indem die informationsverarbeitende Elektronik auf das Potential des
Zwischenkreises (≈ -280 V) gelegt wurde.
Wenn sich ein Fahrzeug von einem Statorsegment zum nächsten bewegt, wandert dann die
gesamte Regelungsaufgabe physikalisch in die nächste Informationsverarbeitungseinheit (DSP).
Ein Datenaustauschbedarf mit der Zykluszeit der Stromregelung (100 μs) entsteht dann zwischen
den benachbarten Modulen. In einem beliebig skalierbaren System entsteht daher auch ein
beliebig großer Kommunikationsbedarf. Dieser Bedarf wird nur mit Hilfe einer direkten (Punkt-
zu-Punkt) Verbindung zwischen den benachbarten Modulen erfüllt. Diese Verbindung wurde hier
mit dem kostengünstigen RS485 Protokoll realisiert.
Ein Leitrechner sorgt dann für die Erzeugung von Positions-Sollwerten für die Fahrzeuge, unter
Berücksichtigung eines Fahrplanes. Die Zykluszeit für die Positions-Sollwerte liegt im Bereich von
1-10 ms. Das Monitoring (Beobachtung von internen Variabeln der zahlreichen Servocontrollern)
wird auch in Echtzeit gewährleistet. Offline sind Download- und Upload-Aktionen von Firmware
und allgemeine Dateien auch möglich. Dafür wurde für die Kommunikation zwischen dem
Leitrechner und den verteilten Servocontrollern der Ethernetbasierte Feldbus EtherCAT
verwendet.
VI
Innerhalb der Bearbeitungsstationen des Systems spielen die Positionierungsgenauigkeit und die
Dynamik eine wichtige Rolle. Deswegen werden innerhalb dieser Stationen Positionssensoren
verwendet. Für den Transport der Materialien außerhalb dieser Bereiche, wurde eine EMK-
basierte sensorlose Regelung implementiert. Dadurch lassen sich weitere Systemkosten sparen.
Im Rahmen dieser Arbeit wurde ein kleiner Ausschnitt eines solchen modularen und beliebig
skalierbaren Systems als Versuchsaufbau realisiert und dessen Funktionsfähigkeit nachgewiesen.
VII
Contents
PREFACE.................................................................................................................................I
ABSTRACT ........................................................................................................................... III
KURZFASSUNG.......................................................................................................................V
CONTENTS .......................................................................................................................... VII
LIST OF SYMBOLS AND ABBREVIATIONS ..............................................................................IX
1. INTRODUCTION .............................................................................................................. 1
1.1. State of the art ........................................................................................................................... 3
1.1.1. Linear electric motors for public transportation ........................................................... 3
1.1.2. Linear electric motors in industrial applications for material handling and factory
automation ......................................................................................................................................... 5
1.2. Starting point and purpose of the work ................................................................................ 11
2. CONTROL STRATEGY..................................................................................................... 13
2.1. Mathematical model of the PMLSM....................................................................................... 13
2.2. Field oriented control and modulation method ................................................................... 17
2.3. Hard real-time communication demands between the distributed controllers ............... 21
2.4. Motion coordination and monitoring.................................................................................... 23
3. POINT-TO-POINT CONNECTION BETWEEN THE DISTRIBUTED CONTROLLERS................ 27
3.1. SPI data transfer ...................................................................................................................... 27
3.2. Communication protocol ........................................................................................................ 29
3.3. Data transmission at the physical layer ................................................................................ 34
3.3.1. Idle-bus state.................................................................................................................... 34
3.3.2. Common-mode voltage and multicommutation problems ......................................... 35
4. INDUSTRIAL FIELDBUS SOLUTION FOR MOTION COORDINATION AND MONITORING .... 43
4.1. Overview of the main real time Ethernet (RTE)-based fieldbuses ...................................... 43
4.2. Data transfer modes over EtherCAT ...................................................................................... 45
VIII
4.3. The EtherCAT master application.......................................................................................... 48
4.4. Software configuration for the embedded servo-controllers ............................................. 50
5. EXPERIMENTAL SET-UP................................................................................................ 52
5.1. The Linear Motor Unit ............................................................................................................ 52
5.2. The Power Processing Unit..................................................................................................... 55
5.2.1. The Line Connection Unit ............................................................................................... 55
5.2.2. The Rectifier Unit ............................................................................................................ 55
5.2.3. The Voltage Source Inverter .......................................................................................... 56
5.2.3.1. Bootstrap circuit..........................................................................................................................58
5.2.3.2. Current measurement and over-current protection ..............................................................59
5.2.3.3. Dimensioning of the DC-link capacitors...................................................................................64
5.2.3.4. The interface with the Distributed Control Unit and the electronic supply .......................65
5.3. The Information Processing Unit........................................................................................... 66
5.3.1. The Distributed Control Unit.......................................................................................... 66
5.3.1.1. The interface with the position sensor.....................................................................................68
5.3.1.2. The interface with the adjacent Distributed Control Units ...................................................68
5.3.1.3. The interface with the Central Control Unit............................................................................69
5.3.2. The Central Control Unit ................................................................................................ 72
5.4. The Position Sensor Unit ........................................................................................................ 72
6. CONTROL RESULTS ....................................................................................................... 75
6.1. Sensor based control ............................................................................................................... 75
6.1.1. Characteristic of the induced voltages .......................................................................... 75
6.1.2. Position control of the vehicle ....................................................................................... 76
6.2. Sensorless, EMF-based control ............................................................................................... 81
6.2.1. Dead-time effect and compensation.............................................................................. 82
6.2.2. The EMF observer ............................................................................................................ 86
6.2.3. The speed and position observer ................................................................................... 88
7. CONCLUSIONS............................................................................................................... 93
APPENDIX............................................................................................................................ 95
BIBLIOGRAPHY ...................................................................................................................100
CURRICULUM VITAE...........................................................................................................105
IX
List of symbols and abbreviations
Symbols:
Voltages:
1 2 3, ,S S Su u u Phase voltages of a stator segment ,Sa Sbu u Stator voltages represented in the stator-oriented coordinate system (ab) ,Sd Squ u Stator voltages represented in the rotor-oriented coordinate system (dq)
1 2 3, ,S S Se e e Induced voltages in the phases of a stator segment ,Sa Sbe e Induced voltages represented in the (ab) coordinate system ,Sd Sqe e Induced voltages represented in the (dq) coordinate system
10 20 30, ,u u u Output voltages of the inverter
ofsu Offset voltage to be added to the reference phase voltages
0Su Star point potential of a stator segment
,CC surgeu Blocking voltage of an IGBT during commutation
lineu Transient voltage on the common negative DC-link bus
CMu Common-mode voltage ,A Bu u Potentials of the RS485 pair of wires for differential transmission of a signal
1 2 3, ,C C Cu u u Compensated phase voltages due to the inverter’s characteristics ,SaC SbCu u Compensated voltage reference values represented in the (ab) coordinate system
0 7,U U Zero-voltage vectors of the inverter
DU Voltage of the DC-link
CapU Voltage on an electrolytic capacitor of the DC-link
diU On-state voltage of the inverter’s freewheeling diode
tiU On-state voltage of the inverter’s IGBT
dtU Average on-state voltage in a control cycle of both diode and IGBT
DVU Deviation from the reference value of the inverter’s output voltage
Currents and Fluxes:
1 2 3, ,S S Si i i Phase currents of a stator segment ,Sa Sbi i Stator currents represented in the stator-oriented coordinate system (ab) ,Sd Sqi i Stator currents represented in the rotor-oriented coordinate system (dq)
List of symbols and abbreviations X
Ci Current through the DC-link capacitors of an inverter
,D aci AC component of the rectifier’s output current
aci AC component of the inverter’s input current
linei DC-link current between two adjacent inverters
tCI Current peak value during the charge time of the local DC-link capacitor
tDCI Current peak value during the discharge time of the local DC-link capacitor
NI Output current of the inverter
1 2 3, ,S S S Flux linkages of the stator phases
PM Flux linkage of the permanent magnet with the stator
Times and Frequencies:
,Sd SqT T Time constants of the current control loop
0T Sampling time
MT Time constant of the motor’s transfer function
Parameters and Gains:
SR Winding resistance of a stator phase
SL Winding inductance of a stator phase ,Sd SqL L Stator winding inductance components in the (dq) coordinate system
vM Mass of the vehicle
MV Gain of the motor’s transfer function
RIV Gain of the current controller’s transfer function
ELC DC-link capacitance of an inverter
,commonL Common stray inductance of the DC-link
, phL Stray inductance in an inverter leg
,eff acL Inductance value of the DC-link bus bar system at high frequencies
,bar acL Self-inductance of a DC-link bus bar at high frequencies
,mt acL Mutual inductance in the DC-link bus bar system at high frequencies
SL Stray inductance between the electrolytic and snubber capacitors
ESRL Equivalent series inductance of the electrolytic capacitor
ESRR Equivalent series resistance of the electrolytic capacitor
SNL Equivalent series inductance of the snubber capacitor
SNC Value of the snubber capacitor
IPML Stray inductance of the IPM connection pins
11 22 11 22, , ,h h g g Gains of the EMF observer
1 2 3 4, , ,p p p p Roots of a 4th order polynomial
1 2 3, ,k k k Gains of the mechanical observer
Cb Speed-dependent friction coefficient
List of symbols and abbreviations XI
ET Time constant of the inverter’s transfer function
RIT Time constant of the current controllers
EIT Equivalent time constant of the current control loops
osct Oscillation time of the current measurement circuit
Ct DC-link capacitor charge time
DCt DC-link capacitor discharge time
DEADt Dead-time
, ,,C ON C OFFt t Turn-on and turn-off times of the IGBTs
ont Time inside a control cycle during which the low-side IGBT is on
1 10...T T Time intervals corresponding to the SPI communication
0 Natural frequency of the current control loop
Nf Output frequency of the inverter
mainsf Frequency of the mains
Gf Output frequency of the bridge rectifier
Angles:
Electrical angle between the phases of a stator Electrical angle of the rotor (vehicle) Difference between the estimated and the real value of the electrical angle
Angle of the inverter’s voltage vector
Subscripts and marks:
minO Minimal value
maxO Maximal value
nomO Nominal value
refO Reference value
RMSO Root mean square value
INO Input value
OUTO Output value O Sum value of the induced voltages in two adjacent stator segments
EO Output angle or position values from the EMF observer
MO Output angle or position values from the mechanical observer
O Estimated value O Estimation error
O Time derivative *O Complex conjugate
O Space vector treated as a complex quantity
List of symbols and abbreviations XII
Matrixes in state space representation:
x Matrix of the state variables A, Ad, Av State matrixes B Input matrix K Disturbance matrix d Matrix of the disturbance variables
2 3I , I Unity matrixes H, G Gain matrixes
Other Symbols:
ga Air gap between the stator and the permanent magnet
rec Recoil permeability
0 Vacuum permeability Pole pitch of the motor v Speed of the vehicle
fv Filtered speed of the vehicle x Distance
ip Instantaneous power at the motor terminals
magp Instantaneous power in the magnetic field of the stator
mechp Instantaneous mechanical power
ohmp Instantaneous value of the ohmic power losses
relp Instantaneous value for the reluctance component of the mechanical power
Real part of an expression j Imaginary unit
thF Thrust force
dF Disturbance force
LF Load force
Fk Force constant s Laplace operator
0D Damping of the control loop
0 ,G K Transfer functions ,pd pq Perturbances of the current control loops
0 21...
, , ,
, ,C D R S
X Y Z
P P
P P P P
P P P
Specific positions of the vehicle along the track of the experimental set-up,
which trigger or define specific actions of the control stage
List of symbols and abbreviations XIII
Abbreviations:
ADC Analog to Digital Converter AMR Anisotropic Magneto-Resistiv ASCII American Standard Code for Information Interchange ASIC Application Specific Integrated Circuit CCU Central Control Unit CLK Clock COFF Common Object File Format CSMA Carrier Sense Multiple Access DCU Distributed Control Unit DSP Digital Signal Processor EtherCAT Ethernet for Control Automation Technology EMC Electro-Magnetic Compatibility EMF Electro-Motive Force EMI Electro-Magnetic Interference ESC EtherCAT Slave Controller EVM Event Manager FMMU Field Memory Management Unit FoE File access over EtherCAT FO Fault Output FOC Field-Oriented Control FPGA Field-Programmable Gate Array FTP File Transfer Protocol GPIO General Purpose Input Output HTTP Hypertext Transfer Protocol HVIC High Voltage Integrated Circuit IGBT Insulated Gate Bipolar Transistor IP Internet Protocol
IPM Intelligent Power Module IPU Information Processing Unit IRT Isochronous Real Time LBM Linear Brushless Motor LCU Line Connection Unit LIM Linear Induction Motor LMU Linear Motor Unit LSM Linear Synchronous Motor LVIC Low Voltage Integrated Circuit LVTTL Low Voltage Transistor-Transistor Logic Mbps Megabits per second
List of symbols and abbreviations XIV
McBSP Multichannel Buffered Serial Port MF Medium Frequency MISO Master Input Slave Output MOSFET Metal Oxide Semiconductor Field Effect Transistor MOSI Master Output Slave Input OSI Open System Interconnection PCB Printed Circuit Board PDPINT Power Drive Protection Interrupt PDI Process Data Interface PE Protection Earth PI Proportional Integrative PLC Programmable Logic Controller PM Permanent Magnet PMLSM Permanent Magnet Linear Synchronous Motor PPU Power Processing Unit PSU Position Sensor Unit PT1 First-order lag element PWM Pulse-Width Modulation RAM Random Access Memory RTE Real Time Ethernet RU Rectifier Unit SPI Serial Peripheral Interface SS Slave Select TCP Transmission Control Protocol TDMA Time Division Multiple Access TTL Transistor-Transistor Logic TwinCAT The Windows Control and Automation Technology UDP User Datagram Protocol VSI Voltage Source Inverter XINTF External Interface WKC Working Counter
1
1. Introduction
The linear electric motor works after similar principles as the rotary electric motor, consisting of
a stator and a mover separated by an air gap. The magnetic fields of these two parts will interact
to produce a linear direct motion, meaning that no mechanical transmission is required between
the stator and the mover. As they use the similar principles for generating the electromagnetic
force, there are as many types of linear motors as rotary motors. Almost every rotary motor type
will find its correspondent in the class of the linear motors. If a rotary motor is unrolled, it will
form a flat, single sided linear motor. Rolling again this flat linear motor about the axis of the
direction of movement will form a tubular linear motor. The magnetic flux of a linear motor can
be parallel to the direction of movement (longitudinal flux motor) or perpendicular to this
direction (transverse flux motor). A combination of the two types of fluxes inside one linear motor
is also possible, where the flux is transverse in the stator and longitudinal in the mover [1]. The
part of the linear motor, which produces the magnetic field in the air gap, is called excitation part
and the other part where voltage is induced is called armature. Each part can be on the stator or
on the mover side. According to the principle of generation of the electromagnetic force, the
linear motors can be divided into three main categories:
Linear Induction Motor (LIM). The part of the motor, which produces a travelling magnetic
field in the air gap, is called primary (excitation part) and consists of three-phase
windings. This travelling field will induce a voltage in the secondary of the machine if
there is a speed difference (slip) between them. The secondary consists of a conducting
plate on a solid iron or of short-circuited three-phase windings placed in slots of laminated
core [2]. The induced voltage will generate currents in the secondary, which are
interacting with the travelling magnetic field to produce the electromagnetic force.
Linear Synchronous Motor (LSM). The armature creates in this case the travelling
magnetic field and consists of three-phase windings. The excitation part can create a
magnetic field or a variable magnetic reluctance in the air gap. This magnetic field can be
produced by a PM or by a DC-current winding and it will interact (synchronize) with the
travelling magnetic field to generate electromagnetic force. In the case of the variable
magnetic reluctance motors, the electromagnetic force is a reaction force of the mover to
the travelling magnetic field.
Introduction 2
Linear Brushless Motor (LBM). This motor works with a rectangular three-phase current
system, whereas the first two above mentioned types are using sinusoidal currents. The
induced voltages in the armature are trapezoidal. Using a position sensor, the
commutation process of the currents is synchronized with the position of the mover. The
LBM is therefore a special type of LSM.
In Figure 1.1 a brief classification of linear motor types is shown, regarding both the motor layout
and the principle of operation. In the case of long-stator LSMs, the armature is longer than the
excitation part. For short-stator LSMs the excitation part is longer than or equal to the armature.
Linear Electric Motor(layout)
Flat geometry Tubular geometry
Longitudinal flux Transverse flux
Double sided Single sided
Slotted Slotless
Long statorLong primary
Short statorShort primary
Linear Electric Motor(operation)
Induction Synchronous Brushless DC
PM DC-currentwindings
Variable reluctance
Figure 1.1 Classification of the linear motors
Introduction 3
For LIMs the terms of long-primary and short-primary are preferred instead. Almost any
combination between layout and operation forms is possible depending on the application.
The linear motors were first mentioned at the end of the 19th century and at the beginning of the
20th century their main applications were meant for high-speed propulsion and public
transportation systems. In 1946 the Westinghouse Corporation builds the first large-scale linear
motor for aircraft launching, called Electropult, which was capable of speeds of about 100 m/s [1].
Even though the idea of using electromagnetic levitation in public transportation systems dates
back in 1922, the first practical construction only took place in 1969 in Munich, Germany and the
model was called Transrapid 01, which was capable of speeds up to 100 km/h [3]. The maglev
(magnetic levitation) transportation system combines the linear motor technology with the
technology of contactless magnetic suspension, offering energy efficient and pollution free
alternative to air, car and standard railway traffic. Starting with the early 1970s the maglev
systems for public transportation developed continuously, as they still are in the present. For
industrial applications like machining, material processing or laser cutting, the linear electric
motors started to win recognition and to be implemented only at the beginning of the 1990s [3].
The continuous progress over the last years in areas like motor design, power electronics, control
of drives, positioning sensors and industrial fieldbuses, together with the actual interest for new
intelligent solutions in transportation and industrial plants, opened new perspectives for the
linear direct drives applications.
1.1. State of the art
1.1.1. Linear electric motors for public transportation
For high-speed (over 400 km/h) maglev systems, the long-stator LSM is preferred. The short-
stator LSM is not used because there is the problem supplying the vehicle with energy and the
vehicle would also be heavy, as all the necessary power converting units are situated on it. The
LIM is also not an alternative for high-speed transportation, because its efficiency is strongly
affected by the wide air-gap and its fluctuations. Only the currents in the primary produce the
thrust force of the LIM, as the secondary has no independent magnetic field. The primary of the
LIM will have therefore bigger dimensions than the armature of an LSM for the same thrust force,
as it must carry the magnetization current component in addition.
In the last years, low-speed (100-150 km/h) linear motor systems become interesting for
passenger transportation in urban areas [4], [6], [7] and also for goods transportation over short
distances inside crowded industrial areas [4]. These transportation systems are with or without
magnetic levitation and for propulsion usually long-stator LSMs or short-primary LIMs are used
[5]. Nowadays ropeless elevators, based on linear electric motors, represent an alternative to the
conventional elevators in very high buildings. The fixed parts of the motors are situated on the
shaft’s wall and the moving parts are the cabins. No drive units situated at the top of the building
are required. More cabins are allowed to move independently on the same shaft and horizontal
Introduction 4
movements are also possible in order to change the shaft. This increases the transportation
efficiency, so that the number of shafts can be reduced. The building’s size and automatically its
construction costs can be therefore reduced. The Multi Mobile System (MMS), developed in
Switzerland and supported by the elevator manufacturer Schindler AG, uses a double-sided, long-
stator LSM with excitation realized by PMs. One stator segment is made up of many motor
modules connected in parallel. An inverter supplies each segment and all the inverters are
sharing the same DC-link bus [8]. For these applications the double-sided, long-stator PMLSM with
a slotted iron core is the best solution, as it develops the biggest thrust-force density compared
with other linear motor topologies [9].
The most popular linear electric motor used now in public transportation is therefore the long-
stator LSM. Because the vehicles travel over long distances, the armature along the guideway is
split in many segments in order to improve the efficiency [10]. Only those segments involved in
the development of the thrust force are energized. The segments have different lengths and can
be separated by different gap sizes. In systems requiring relative high throughput (many
travelling vehicles per time unit), as in the case of elevator systems, the segments are small
compared with the size of the vehicle and are supplied by dedicated inverters.
The throughput is low for large distance transportation systems, so that the segments are long
and a set of multiplexed inverters supplies several segments by switchgears. Different inverters
can also supply two adjacent segments. When the vehicle passes over the respective segments,
both inverters must be synchronized so that the desired thrust force is generated. A central
control unit is responsible for the position and speed of the vehicle. These actual values are
transmitted over a wireless communication system from the vehicle to the control units. The
distributed control units manage the power flow from the inverts to the respective segments
according to new position and speed reference values generated by a traffic planner.
The system topology of the Transrapid system in Shanghai is shown in Fig. 1.2. It is a modular
control system, where the modules, called substations, are located along the guideway [11].
DU
DCU
LCU
DCU
DU
LCU
SCU SCU SCU
CCU
CCU = Central Control UnitDCU = Distributed Control UnitDU = Drive UnitLCU = Power-Line Connection UnitSCU = Switch Control Unit
Power LinesFiber Optic Bus (OTN Network)Redundant Communication BusRedundant Radio Communication
Substation „n“ Substation „n+1“
Motor SectionSwitched on
Figure 1.2: Topology of the Transrapid system in Shanghai
Introduction 5
1.1.2. Linear electric motors in industrial applications for material handling and factory automation
Linear electric motors are used nowadays in industrial applications such as semiconductor
industry, flat panel display manufacturing, machine tools, electronics assembly, factory and
laboratory automation, robotics and photovoltaic cell manufacturing. Compared to conventional
systems using pulleys, racks, screws or other mechanical parts that transform rotary to linear
motion, direct drive systems based on linear motors have higher processing precision and lower
maintenance costs. The implementation costs can be higher, but they are compensated in time by
the higher throughput capability.
Material handling is an emerging application, which benefits from the advantages of the linear
drive systems. Here, small to medium sized materials must be positioned with high dynamics
between special stations, where they have to be processed with a high degree of accuracy. A single
linear direct drive system is used for both transportation and processing tasks. Inside this class of
applications there are several possible topologies [12], depending on specific features of the
application, like the size and profile of the production track, the number of material carriers
(vehicles), the dynamic requirements, the power rate and the implementation costs. Typical
specifications in material handling and factory automation applications are [13]:
Positioning accuracy around 50 μm
Repeatability 10-20 μm
Demanding duty cycles
High dynamics and reliability over cost ratio
The general system configuration for this type of applications is shown in Fig. 1.3. The main units
of the system are:
Linear motor unit (LMU)
Power processing unit (PPU)
Information processing unit (IPU)
Position sensor unit (PSU)
At the LMU-level the PMLSM is used for its high thrust-force density and for the fact that air-gaps
of few millimetres are allowed. Both short-stator and long-stator PMLSM types are used. The iron
core can be slotted or slotless. The slotted type has a bigger thrust force per volume unit, but has
IPUIPC, DSPμC, FPGA
PPU LMU
PSU
Mains
Net-to-DCDC-Link
DC-to-AC PMLSM
OpticMagneticCapacitiv
3x 400V50Hz (Net)
Electrically IsolatedSignal Transfer
Power Lines
Noise InsensitiveSignal Transfer
Figure 1.3: Configuration of the direct drive system
Introduction 6
therefore a significant cogging force. The most popular technique to reduce the cogging effect is
skewing the magnets [3]. The single-sided configuration of the PMLSM usually fulfils the dynamic
requirements, but in the case of high acceleration demands, or if the attractive force between the
PM and the iron core is too high, the double-sided configuration can be implemented. Rare-earth
magnets such as NdFeB (Neodymium-Iron-Boron) or SmCo (Samarium-Cobalt) are selected for the
excitation part of the LSM due to their high remanence and energy density values.
At the PPU-level a standard voltage source inverter (VSI) supplies the windings of the armature.
Multiplexing inverters by mechanical or electronic switches is not a feasible solution for long-
stator topologies in this case. The mechanical switches are not applicable due to the large number
of switching actions per time unit and a safe-functioning electronic switch unit will require a
larger number of IGBTs than using one dedicated inverter for each motor section.
At the IPU-level there is always a central control unit (CCU) responsible for the traffic flow inside
the system. It generates the new position reference values of the vehicles, monitors the entire
process and interacts with the system administrator over a graphical interface. The CCU is
connected with the PPU by means of a controller, such as microcontroller, DSP or FPGA. The
connection between this controller and the PPU is typically electrically isolated, where the pulse
width modulated (PWM) signals generated by the controller are connected to optocouplers and
then transferred to the gate-drive units of the power electronic devices of the VSI (IGBTs or
MOSFETs). The position control algorithm is implemented either in the CCU or in the above-
mentioned controllers. The most popular control method for servo drives is the field-oriented
control (FOC), where all calculations are performed in the orthogonal moving frame of the
vehicle. For this reason, the position of the vehicle must be known. All the controllers inside one
system have to communicate with the central control unit by means of an industrial
communication protocol.
There are three linear direct drive topologies used for material handling and factory automation.
1) Short-stator topology
It is still the most wide spread topology in this class of industrial applications. The track is passive
and contains the PMs and the vehicles are the active part (moving windings). The energy is
brought to the vehicles by cable drag chains or by sliding contacts. The motor system from the
Parker Company, shown in Fig. 1.4, is an example of such topology. The PPU contains a standard
VSI and at the IPU-level there is a controller designated to each vehicle. The controller
communicates with the CCU over a cable or a wireless connection. The controller, the VSI and the
position sensor are physically situated on the vehicle. Due to the complexity of the cable structure
module, it is very difficult to use this topology over long distances or if many vehicles are present
in the system. The difficulty due to long and curved tracks can be avoided by using a contact-less
energy transmission at medium frequency (MF), as shown in Fig. 1.5. In this case the PPU inside
one vehicle contains beside the VSI also the secondary of the MF-transformer and a rectifier. Due
to the fixed power rate of this transformer, if many vehicles are simultaneous in acceleration or
Introduction 7
deceleration phases, this alone cannot assure the dynamic requirements. For this reason, the
short-stator topology is only recommended for systems with short tracks, high dynamics and low
number of vehicles or for systems with long tracks and low dynamics. The dynamic performance
can be improved if rechargeable power supplying elements like super capacitors are used on the
board of the vehicles. These elements are discharged during acceleration periods and are again
recharged during braking. They are also a part of the vehicle’s PPU along with the power
converters that assure this energy flow, leading to relative heavy and complex vehicle structures.
There are research projects focusing on this topology [15].
2) Centralized long-stator topology
In some industrial plants for material handling, beside the high processing accuracy, a high
throughput and a high complexity of the production track are also required. It involves a high
number of vehicles per track length, which have to move with a high degree of independency,
meaning that they are all able to accelerate and brake simultaneously without loosing their
dynamic. The track can have curves, close paths and switches. For these considerations, the long-
stator PMLSM topology is the suited one. The track is spilt into several armatures (segments) and
only one vehicle may be present on a segment at a given time. The vehicles are passive elements
(moving magnets). Typical segment lengths are between 0.5 m and 1 m. A dedicated inverter
1. Pass-through cabling2. Connector panel
3. High-strength aluminum body4. Magnet single rail5. Slotless linear motor
6. Linear guidance system7. Integral linear encoder8. Limit/Reference sensors
9. Cable transport module10. Protective seals
Figure 1.4 Short-stator PMLSM with drag chains (Parker) [14]
125
6
78 3
9 11
1. Central Control Unit 2. Mains3. MF Power Generator
4. MF Transmission Line5. Distributed Controller Unit6. Voltage Source Inverter
7. MF Transformer 8. Sup. Capacitor & DC/DC Converter9. Slotted Linear Motor
10. Magnetic Single Rail11. Vehicle12. Wireless Transceiver
4
10
12
Figure 1.5 Short-stator topology with contact-less energy transmission
Introduction 8
supplies each segment. In the case of a centralized long-stator topology both the PPU and the IPU
are centralized, e.g. situated in a cabinet.
The implementation of a direct drive system based on this topology is supported by FVA
(Forschungsvereinigung Antriebstechnik), a network for innovative drives in Germany [16]. The
system is quite complex due to the structure of the IPU. The structure of the PPU is simple. A
standard VSI is designated to each stator segment and all VSIs are connected to the same DC-link
rails. The algorithm for current, speed and position control is executed inside the CCU at cycles of
100 s. This cycle time is split in ten intervals, each interval being dedicated to one vehicle. The
prototype system can have therefore a maximum of ten vehicles. The CCU is made of one or more
industrial PCs operating with a real-time Linux kernel. An inverter-bus is used for data transfer
(modulation information and measured phase currents) between the IPU and the PPU. Two
controller-based interfaces are needed; one to connect the CCU to the inverter-bus and the other
to connect the VSIs to the same bus, as shown in Fig. 1.6.
The complexity of the CCU and the number of the vehicle-interface units increases along with the
number of vehicles. The bandwidth of even the most powerful industrial filedbuses will be fully
occupied in this case, so that the inverter-bus will always represent a bottleneck for the scalability
of this system. The motor-supply cables represent another disadvantage of this topology,
especially for large systems. They increase the overall cost, occupy a lot of space and can be very
long. Long cables between the inverter and the motor will cause dangerous overvoltages at the
motor terminals. At the same time the electromagnetic interference (EMI) generated by the
inverter increases [17].
This topology is therefore suitable for small-to-medium systems with no more than ten vehicles.
3) Decentralized long-stator topology
The industrial applications tend to be more complex and larger, because the market demands for
quality and quantity are constantly increasing. Therefore the need to process a bigger quantity of
CCU
VI1
VI2
VI9
VI10
Inve
rter
-Bu
s
PI1
PI2
PI3
PIn
PIn-1
PIn-2
VSI1
VSI2
VSI3
VSIn-2
VSIn-1
VSIn
Cabinet V10 V2V7
CCU = Central Control UnitLCU = Line Connection Unit PIx = Power Interface UnitRU = Rectifier Unit
VIx = Vehicle Interface UnitVSIx = Voltage Source InverterVx = Vehicle number
1 2 3 n-2 n-1 n
RU MainsLCU
Figure 1.6 Long-stator topology with centralized IPU and PPU
Introduction 9
products in a shorter time is high. Due to this trend, the long-stator topology for material
handling applications became interesting in the last years. The centralized topology, as shown,
limits the capability of a system to grow. A decentralized long-stator topology, as proposed in this
work, is a solution to this problem. The structure of a decentralized system is shown in Fig. 1.7.
In this case one stator segment is also supplied by one VSI. The algorithm for current, speed and
position control is entirely executed in the main controller of the DCU. The DCU also contains an
interface that connects the unit to an industrial fieldbus. The CCU contains an industrial or a
standard PC and a fieldbus interface. When a vehicle passes over two adjacent segments, the
respective inverters must be synchronized, so that the desired thrust force can be generated at
the vehicle. For the synchronisation, the involved DCUs have to exchange data within each
current control cycle. This data transfer cannot be realized by means of an industrial fieldbus,
because the same bottleneck problem will occur as in the case of a centralized topology. A special
data bus connection between the adjacent DCUs has to be used instead, so that the entire
bandwidth of the fieldbus serves only for monitoring and traffic control. The size of the system
can be therefore considerably increased. Another advantage of this topology is its high degree of
modularity. One module consists of a stator segment, the VSI and the DCU. Modules can be easily
replaced or added to the system.
This type of topology is relative new and at the moment there are not many such systems
available on the market or under research. One of the companies who developed a similar system
is MagneMotion (USA). The central part of their system, presented in Fig. 1.8, is the QuickStick
module, which can be 0.5 m or 1 m long and contains one or more stator segments (maximal ten),
the necessary VSI and DCU and a position sensor system as well. The adjacent modules are
connected over serial communication lines. A PC or PLC represents the CCU and the industrial
fieldbus used here is EthernetIP. The DCUs are connected to this fieldbus over node controllers,
which act as switches at junction points (turntables or electromagnetic switches) where more
than two paths meet [18].
CCU
1
CCU = Central Control UnitDCUx = Distributed Controller UnitLCU = Line Connection UnitRU = Rectifier UnitVSIx = Voltage Source Inverter
2 3 4 n-3 n-2 nn-1
VSI1 VSI2 VSI3 VSI4 VSIn-3
VSInVSIn-2
VSIn-1
DCU1 DCU2 DCU3 DCU4 DCUn-3
DCUnDCUn-2
DCUn-1
Motor supply linesDC-link linesInverter control signalsData bus between adjacent DCUsStandard industrial fieldbus
RULCU
Mains
Figure 1.7 Long-stator topology with decentralized IPU and PPU
Introduction 10
The advantages of the topology in Fig. 1.7 are obvious. But at the first glance it also seems to be
expensive due to the large number of controllers required. Investigations are still necessary in
order to find solutions that reduce the implementation costs and the overall system complexity.
Finally, the table below is a review of the characteristics of the three topologies presented above.
Figure 1.8 MagneMotion’s modular system structure [19]
System topology: Short-stator Long-stator
System characteristics:
Cable drag chains
Contact less energy transmission
Centralized
control
Distributed control
PPU
Low (one VSI per
vehicle)
Medium (without Sup. Capacitors) High (with Sup.
Capacitors)
Medium (one VSI
per stator segment)
Medium (one VSI
per stator segment)
IPU Low (one controller per vehicle) High
Complexity
PSU Low (sensor chip on the vehicle) High (sensor chip along the track)
PPU Low Medium High
IPU Low Medium-High High
Costs (Over track length unit)
PSU Low (for a certain accuracy level) Medium (for the same accuracy level)
Dynamic
High
Low (without Sup. Capacitors)
Medium (with Sup. Capacitors)
High
Scalability Low Medium Medium High
Track length Very short Long Medium Long
Nr. of vehicles Very small Small Medium High
Type of
application
Throughput Medium Medium High High
Table 1.1: Direct drive system topologies for industrial applications based on PMLSM
Introduction 11
1.2. Starting point and purpose of the work
The starting point of this work was fixed on decentralized long-stator topology systems (Fig. 1.7),
based on PMLSM, for industrial applications like material handling and factory automation. The
purpose was to design and implement such a system, where the costs and complexity had to be
reduced as much as possible without affecting the specifications mentioned in the section 1.1.2.
The issues that appeared under these circumstances at the PPU and the IPU level had to identified
and solved. The key aspects of this work are:
The development of the control strategy (chapter 2). The sensor-based, as well as the
sensorless EMF-based control were considered. The sensorless control is an important
issue for long-stator topologies, where the position-sensing unit is very expensive.
The development of a point-to-point communication structure between the DCUs with a
minimal hardware and software complexity (chapter 3).
The analysis and selection of the most suitable industrial fieldbus for this application, as
well as its integration in the structure of the IPU (chapter 4).
The development of a cost-effective servo-controller (VSI + DCU), where the ground
potential of its electronic is tied at the negative DC-link rail (about –280 V). All servo-
controllers are connected then to the same DC-link bus (chapter 5).
The test of the developed system and of the control strategy, by means of a small-scale
experimental set-up. This is meant to confirm the high scalability, the reliability and
dynamic characteristics of the developed architecture (chapter 6).
13
2. Control strategy
This chapter presents the theoretical aspects regarding the control of the system. The adopted
control strategy has to fulfil the specifications presented in the previous chapter regarding
position accuracy, repeatability, throughput, reliability and dynamics. The FOC algorithm is based
on the mathematical model describing the fundamental wave behaviour of the PMLSM without
saturation and takes into account the synchronization requirements of two adjacent inverters
during the vehicle’s transition period across the respective stator segments. The algorithm is
entirely implemented in the DCU where it is executed cyclically each 100 μs. The synchronization
requires a data transfer in each cycle between the adjacent controllers. This represents the hard
real-time communication demand of the control system. The motion coordination of the vehicles
and the monitoring of the system is the task of the CCU. According to its traffic planner, the CCU
must send cyclically new position reference values to the DCUs and collect status data from them.
This data transfer has a cycle time of few milliseconds and represents the soft real-time
communication demand of the control system.
2.1. Mathematical model of the PMLSM
Based on equations presenting the relations between electrical and magnetic values in the motor,
the mathematical model describes the generation of the forces acting on the vehicle. The model is
valid for both long-stator and short-stator topologies and considers no saturation and no space
harmonics. The motor excitation is realised by rare earth PMs, which are characterized by a high
coercitive magnetic field strength and low recoil permeability 0rec . They are subject to
demagnetisation due to the reluctance of the entire magnetic circuit and the stator’s reaction
field, but are not affected by fluctuations in the air gap and can withstand strong demagnetisation
C+
A-
B+
C-
A+
B-
C+
A-
B+
C-
A+
B-
C+
A-
B+
C-
A+
B-
C-
A+
B-
C+
A-
B+
Laminated iron coreSegment nSegment n-1 Segment n+1
τ
Solid back iron
N Sτ τ
Vehicle with skewed PMs
ag
Fundamental wave of the fluxdistribution of the PMs
Fundamental wave of theinduced EMF
Figure 2.1 Structure of the segmented, long-stator PMLSM
Control strategy 14
cycles. For the PMLSM used in the experimental set-up, the PMs are bounded on the surface of a
solid back iron and the stator windings are distributed in slots of a laminated iron core as shown
in Fig. 2.1. The space distribution of the flux density in the air gap, produced by both PM and
stator currents, for a motor topology similar to the one in Fig. 2.1, was determined in [20]. Only
the fundamental of this value is considered, so that the EMF in the motor is assumed to be
sinusoidal. The phase voltage equations of the PMLSM are:
The equation (2.1) in fixed stator coordinates can be written by means of space vectors as in [21]:
where
Remark: For the definition of space vectors, different scaling factors can be used. In this thesis,
the scaling factor 1k is used, as in [21]. Frequently, in other textbooks 2 / 3k is used. With
2 / 3k the projection of a space vector to a winding axis of the machine directly gives the
instantaneous value of the physical quantity ( , ,i u etc.). Using 1k the physical quantity is
obtained by multiplying the projection of the space vector by 2 / 3 .
Similar equations to (2.3) are also valid for the current and flux space vectors. The electrical angle
between the stator phases is 2 / 3 . The star connection of the motor and the sinusoidal EMF
leads to the following constraints regarding the phase currents and the linkage fluxes:
The space vector of the stator flux linkage is defined as:
PM is the position dependent value of the PM flux linkage with the stator, SL is the stator
inductance and is the electrical angle of the vehicle.
11 1
22 2
33 3
SS S S
SS S S
SS S S
d tu t R i t
dtd t
u t R i tdt
d tu t R i t
dt
(2.1)
S
S SS
d tu t R i t
dt
(2.2)
21 2 3
j jS S S Su t k u t u t e u t e (2.3)
1 2 3
1 2 3
0
0
S S S
S S S
i t i t i t
t t t
(2.4)
3
2j t
S SS PMt L i t x t e (2.5)
t x t
(2.6)
Control strategy 15
Using (2.5) and (2.2) the real and imaginary components of the voltage space vector in the fix
stator reference frame ab are:
where ( )v t is the speed of the vehicle.
In order to simplify the non-linear equations in (2.7), the voltage space vector will be shifted in a
system of coordinates dq , which follows the magnetic field of the vehicle. The voltage equations
in dq frame become:
Theoretically the EMF exists only in the q-axis and has the value of Sqe . The instantaneous power
at the motor terminals is:
The equation (2.9) can be written in dq reference frame
By inserting (2.8) in (2.10), the instantaneous power is determined by:
This power is split into ohmic power losses ohmp t , power stored in the magnetic field magp t
and the mechanical power mechp t transferred to the vehicle.
3sin ( )
2
3cos ( )
2
SaSa S Sa S PM
SbSb S Sb S PM
di tu t R i t L x t t v t
dtdi t
u t R i t L x t t v tdt
(2.7)
3
2
SdSd S Sd Sd Sq Sq
SqSq S Sq Sq Sd Sd PM
Sq
di tu t R i t L L i t v t
dtdi t
u t R i t L L i t v t x t v tdt
e
(2.8)
1 1 2 2 3 3
*2
3
i S S S S S S
S S
p t u t i t u t i t u t i t
u t i t
(2.9)
*
2
3
S Sd Sq
S Sd Sq
i Sd Sd Sq Sq
u t u t ju t
i t i t ji t
p t u t i t u t i t
(2.10)
2 22 2
3 3
2
3
SqSdi S Sd Sq Sd Sd Sq Sq
ohmmag
Sd Sq Sd Sq PM Sq
rel
m
di tdi tp t R i t i t L i t L i t
dt dtp t p t
L L i t i t v t x t i t v t
p t
p
ech t
(2.11)
Control strategy 16
The reluctance component ( )relp t of the mechanical power is zero if the current Sdi is controlled
to zero or if the motor has no saliency. The reluctance motor, which has no excitation field, uses
only the reluctance component to produce the thrust force. The relation between the mechanical
power and the thrust force is:
If we consider the reluctance force to be zero, the relation of the thrust force is:
The effect of this thrust force on a vehicle of mass vM is:
The characteristics of the PMLSM are summarised in the block diagram of Fig. 2.2.
The force dF , seen by the control system as a perturbation force, is the sum of three forces:
The friction force of the linear guiding system, which depends on the vehicle’s speed [22]
The periodical cogging force, which is produced by the interaction between the PMs and
the slotted stator (which was neglected in the fundamental wave model)
The load force
The cogging force depends on the electrical angle . Skewing the PMs diminishes its amplitude.
The bandwidth of the speed control loop must be big enough in order to compensate the cogging
force, especially at high speeds. Otherwise compensation methods as in [23] would be necessary.
mech thP vF (2.12)
th PM Sq f SqF x i k i
(2.13)
v th d
dvM F F
dt
(2.14)
uSq
uSd
LSd - LSqLSq
LSd
ΨPM
(x)
v x
Fd
Fth
+
-
/
- -+
iSq
iSd
3 / (2 )
Figure 2.2 Block diagram of the PMLSM
Control strategy 17
For this application the current Sdi is controlled to zero, because it doesn’t contribute to the
thrust force generation, as the saliency is negligible small. The cogging force effect of the motor
chosen for the experimental set-up is very small, so that no compensation was necessary.
2.2. Field oriented control and modulation method
The equations describing the behaviour of the PMLSM in dq reference frame are the basis for
the design of the current, speed and position control loops. The variables of these equations have
DC values in steady state and lead to a decoupled control system similar to the one of a DC motor.
The proportional position controller, the PI speed controller and the PI current controller form a
cascaded control system, which is the classical method for the control of electrical drives. The
current control loop is the inner loop and is therefore very important for the dynamic of the
whole system. The design of the dq current controllers starts from equations in (2.8), written as:
are the time constants of the current control loops.
The current controllers for the d- and q-components are running in
parallel and are designed in the same manner. In (2.15), both equations
are non-linear due to the coupling terms pd and pq , which are acting as perturbations for the
control loops. The second term of the component pq , which depends only on the speed, is
considered a stationary perturbation, so that only the current dependent term is important for
the dynamic performance of the controller. At low speeds, the influence of the perturbations can
be very well compensated by the integrative component of the controllers. At high speeds,
depending on the characteristics of the PI controller and of the motor, a good dynamic
performance can be also achieved without using feedforward compensation.
Fig. 2.3 shows the current control loops, where the behaviour of the motor and of the inverter is
1
1 1 3
2
SdSd Sd Sd Sq Sq
S
SqSq Sq Sq Sd Sd PM
S S
di tT i t u t T i t v t
dt Rpd
di tT i t u t T i t v t x t v t
dt R R
pq
(2.15)
SqSdSd Sq
S S
LLT T
R R
uSd,q ref
+
-iSq ref
pd
iSd,quSd,q
pq-Current
Controller Inverter
Motor Model
iSd ref = 0TRI TE
VRI
TM
VM
Figure 2.3 dq current control loops
Control strategy 18
associated with a PT1 element. In (2.16), 0T represents the sampling time (100 μs).
The transfer function of the current controller is:
The controller is designed by using the method of amplitude optimum, where the time constant of
the controller must compensate the time constant of the motor RI MT T , so that the closed-loop
transfer function will describe a second order system:
where 0 is the natural frequency and 0D is the damping of the control loop. From (2.18) the gain
of the controller is determined for the desired damping value. The output values of the current
controllers are limited according to the maximal value of the voltage space vector Su .
The d-component of the voltage will have priority over the q-component, as shown in (2.19).
The value of maxSu depends on the modulation method used. A simple, code efficient way to
design the modulator is to use directly the phase reference voltages 1 2 3, ,S ref S ref S refu u u , generated
as in Fig. 2.8. For a good utilization of the inverter’s voltage limits, the star point 0Su from Fig. 2.5
can be shifted from its zero position by using an offset voltage, as shown in Fig. 2.6. The switching
time for each phase will be:
1, 1
1 ; ,
MLSM
M
M M Sd SqS
VPT
T s
V T T TR
(Motor)
1,
0
1
1
3
2
INVE
E
PTT s
T T
(Inverter)
(2.16)
1( ) RI
RIRI
T sK s V
T s
(2.17)
0
2 2 02
0 0
1 1( )
211 1M E S M S
RI RI
G sT T R T R D
s s s sV V
(2.18)
usq ref
-iSq ref
iSq
TRi
VRi
-+
usq maxiSq ref
iSq
VRi
TRi
usq ref
usq min
Figure 2.4 PI controller with anti-windup
2 2,max,max ,max ,max
2 2,max,min ,min ,max
SSd Sq Sd Sdref
SSd Sq Sd Sdref
u u u u u
u u u u u
(2.19)
0 (Y 1, 2, 3)
1 ;
2SYref ofs
onYD
u ut T
U
(2.20)
Control strategy 19
In (2.20) onYt is the on time of the low side IGBTs (as defined in the list of symbols), DU is the DC-
link voltage and ofsu is the offset voltage.
There are many modulation methods depending on how
many inverter legs are switching in one PWM cycle and
on how the offset voltage is being calculated [24], [25],
[26], [27]. The average value of the modulation index and
the current measurement by shunts are important
factors in selecting the modulation method, as it will be
illustrated in the fifth chapter.
Fig. 2.7 shows the simulated offset voltage for two
carrier based modulation methods. For both methods,
the maximal values of the space vector voltage and of
the phase voltages are:
iS1
~
~
~
RSS1
S2
S3
0
1
1
1
u10UD/2
UD/2
S0
b
a
[1 0 0][0 1 1]
0
0
[0 0 1] [1 0 1]
[0 1 0] [1 1 0]
uS
[S1 S2 S3]
u20
u30
iS2
iS3
uS3
uS2
uS1
uS0
RS
RS
LS
LS
LS
U3
UD
UD3
2
See remark on page 14
*)
*)
*)
U4 U7
U0
U5 U6
U1
U2
Figure 2.5 Equivalent inverter circuit
uS1UD/2
0
uofs max
UD/ 3
S
uofs min
uS2uS3
Figure 2.6 The offset voltage
u10u20 u30uofs u20 u30 u10uofs
UD
3
UD
3
UD
3-UD
3-
Vol
tage
[V]
Vol
tage
[V]
a) Carrier based modulation with the offset
voltage calculated as in (2.22) b) Carrier based modulation with the offset
voltage calculated as in (2.23)
Figure 2.7 Low frequency content of the offset voltage and the inverter’s output voltages
,max ,max (Y 1, 2, 3)
3 ; ;
2 3D
S SYD
Uu U u
(2.21)
Control strategy 20
The offset voltage is calculated according to (2.22) for the modulation of Fig. 2.7a and to (2.23) for
the modulation of Fig. 2.7b ([27], [28]).
For the design of the speed control loop, the current control loop of the q-component is
approximated by a PT1 element.
Fig. 2.8 shows the block diagram of the entire cascaded control. In order to avoid a big speed
overshoot, the speed reference value is filtered by means of a first order low-pass filter. The
design of the speed and position control loops is presented in the appendix.
As it can be seen from Fig. 2.8, the FOC requires the knowledge of the actual position and speed of
the vehicle. A high-resolution sensor can measure the position and then the speed is obtained by
using derivative and filtering operations. If the sensor’s resolution is not satisfactory, a good
max 1 2 3max min
min 1 2 3
max , , ;
2 min , ,
S S ref S ref S refS Sofs
S S ref S ref S ref
u u u uu uu
u u u u
(2.22)
max max min
min max min
; 0 3
; 03
SS S S
ofs
SS S S
uu u u
uu
u u u
(2.23)
20
1( ) ; 4
1I EI EEI
G s T D TT s
(2.24)
uSq refxref t1PWM
VSIM
odu
lato
r
ab
dq 2
3
ab
dq 2
3
isd ref = 0
-
iSq
-
/
or
PMLSM
Speed calculation based onthe sensor's position
Sensorless, EMF based, method
Estimation of the EMF voltagesEstimation of speed and position
- measured position
- estimated position
--
- estimated speed
X
Vf X
- estimated field angle
or
X or X Vf or
x xx
V
vref
V
isq ref
uSd ref
uSa ref
uSb ref
uS1 ref
uS2 ref
uS3 ref
t2
t3
iSd
iSa
iSb
iS1
iS2
iS3
X or XV
X
V
X
iSa
iSb
uSa refuSb ref
Figure 2.8 Block diagram of the cascaded control in field coordinates for the PMLSM
Control strategy 21
speed calculation is still possible by using a speed observer based on the mechanical motor model.
For electrical drives, there is generally nowadays desired to implement sensorless control
strategies in order to reduce costs. And this is even more evident in the case of linear drives,
where the sensor system would have to cover the entire length of the track. There are two main
sensorless control strategies:
EMF-based control strategy
Control strategy based on magnetic anisotropies
The first one doesn’t work at standstill but is relative simple to implement, while the second one
works theoretically at standstill but is more complex and quite challenging especially at high
loads [29], [30], [31], [32]. For the applications that this work focuses upon, the EMF-based method
alone is satisfactory. The implementation of the EMF-based control and the experimental results
for both sensor-based and sensorless control are presented in detail in chapter 6.
2.3. Hard real-time communication demands between the distributed controllers
The block diagram from Fig. 2.8 is implemented in all distributed control units (DCUs) of the
system. During the transition period of a vehicle across two adjacent stator segments, both
involved DCUs have to control the current of the equation (2.13) in such a way that the desired
resulting force is developed at the vehicle, which has to assure a smooth, bump-less transition.
Therefore, a data exchange between the adjacent DCUs must take place within each current
control cycle during the transition period. This period contains three main states:
Segment „n+1“Segment „n“
iS1
isq ref
X^
eSa(n+1)^
Positionand SpeedControllers
d,q CurrentControllers
Phase Transf.
2
PWM
VSI
usa ref
iSb
,
EMF-Observer
„n“
EMF-Observer
„n+1“
Positionand SpeedObserver
„n“
Positionand SpeedObserver
„n+1“
Positionand SpeedControllers
X
V
V
x xx x xx
VSI
DCU “n“ DCU “n+1“
d,q CurrentControllers
Phase Transf.
2
PWM
usb ref
iSa
usa ref
iSb
usb ref
iSa
eSb(n+1)^
i sq
ref
i sd
ref =
0
iS2iS3
iS1
iS2
iS3
Figure 2.9 Data transfer between the DCUs at the beginning of a transition period
Control strategy 22
The first state begins shortly before the vehicle reaches the edge of one segment and is
heading towards the neighbour segment as shown in Fig. 2.9. In this figure the case of an
EMF-based sensorless control is considered. The current controllers of both stator
segments must be active, but only one DCU is controlling the position and the speed of the
vehicle. It is a master-slave relation where DCU “n” is the master and DCU “n+1” is the
slave. DCU “n+1” has to control the current according to the received reference value sent
by DCU “n”. Similarly, only the position and speed observer of the master DCU is active but
this needs the EMFs from both segments. This state ends when the vehicle reaches the
middle point between the stator segments.
The second state begins when the vehicle reaches the central position between two stator
segments and ends after one control cycle. Within one control cycle, the master-slave roles
of DCU “n” and DCU “n+1” interchange, so that DCU “n+1” becomes the master. This is also
the state where the maximum amount of data has to be exchanged. The final values of the
state variables for speed and position control must be transferred from DCU “n” to DCU
“n+1” where they are set as initial values, as shown in Fig. 2.10. For the position and speed
control the state variables are the integrator value of the PI speed controller and the value
of the filtered speed reference. For the position and speed observer, the state variables are
the estimated speed and position values.
The third state follows immediately after the second state and ends shortly after the
vehicle completely leaves segment “n”. Now DCU “n” has to control the current according
Segment „n+1“Segment „n“
Positionand SpeedControllers
d,q CurrentControllers
Phase Transf.
2 3 & ab
PWM
VSI
EMF-Observer
„n“
EMF-Observer
„n+1“
Positionand SpeedObserver
„n“
Positionand SpeedObserver
„n+1“
Positionand SpeedControllers
V
x xx x xx
Transfer of State Variables
d,q CurrentControllers
PWM
VSI
Phase Transf.
2 3 & ab
DCU “n“ DCU “n+1“
usa ref
iSb
usb ref
iSa
usa ref
iSb
usb ref
iSa
isq ref
eSa(n+1)^ eSb(n+1)
^,
X^ X
V
i sq
ref
i sd
ref =
0
iS1iS2
iS3
iS1iS2
iS3
Figure 2.10 Maximum data transfer between two adjacent DCUs during a transition period
Control strategy 23
to the received reference value sent by DCU “n+1”. Also here both EMF-observers must be
active, as shown in Fig. 2.11.
Regarding the master-slave relation at the level of the control strategy, the mastership moves
according to the vehicle’s position and an additional slave control function is invoked while the
vehicle transits to the next stator segment. As an arbitrary number of vehicles may be
simultaneously in a transition period, a high communication bandwidth will be required. For a
highly scalable system any industrial fieldbus will not be able to fulfil this task. The necessary
communication bandwidth can be only guaranteed by individual point-to-point connections
between the adjacent DCUs. Any DCU must be able to communicate with both adjacent units, on
its left and on its right respectively. It must poses therefore two independent and relative fast
communication channels. At the application layer this communication was realized by means of
the SPI (Serial Peripheral Interface) protocol, as the DCU has two SPI compatible units. At the
physical layer the differential data transfer protocol RS485 was used. The implementation of this
point-to-point connection is presented in chapter 3.
2.4. Motion coordination and monitoring
At the beginning of this chapter the main tasks of the Central Control Unit (CCU) were mentioned,
namely: the cyclical data transmission of new position reference values to the DCUs (motion
coordination) and the cyclical status data reading from the DCUs (system monitoring). Inside the
control system we distinguish between two real-time communication loops: the superimposed
loop in milliseconds cycles between the CCU and the DCUs and the hard real-time communication
Segment „n+1“Segment „n“
Positionand SpeedControllers
d,q CurrentControllers
Phase Transf.
2 3 & ab
PWM
VSI
EMF-Observer
„n“
EMF-Observer
„n+1“
Positionand SpeedObserver
„n“
Positionand SpeedObserver
„n+1“
Positionand SpeedControllers
Phase Transf.
2 3 & ab
V
x xx x xx
d,q CurrentControllers
PWM
VSI
Phase Transf.
2 3 & ab
DCU “n“ DCU “n+1“
usa ref
iSb
usb ref
iSa
usa ref
iSb
usb ref
iSa
i sq
ref
i sd
ref =
0
X
VX^
eSa(n+1)^ eSb(n+1)
^,
isq ref
iS2
iS1
iS3
iS2
iS1
iS3
Figure 2.11 Data transfer between the DCUs at the end of a transition period
Control strategy 24
loop executed at every sampling interval of 100 μs between any two adjacent DCUs. Both
communication loops must run independently so that the CCU must be able to access any DCU in
the system, at any moment, without stopping or delaying a running communication between this
and another DCU. Similarly, all DCUs must be able to provide actual status data to the CCU at any
time instant. For this reason, the control system needs another communication bus beside the
SPI-based one. Regarding the scalability of the system, a standard industrial fieldbus is used
therefore for motion coordination and monitoring, which must allow in the first place the
connection of a large number of communication nodes to the control network and a high
communication bandwidth. There are also other requirements, which have to be fulfilled by this
fieldbus, such as:
It must allow high topology flexibility. Line, tree and star connections of the
communication nodes should be possible without additional node controllers or cabling
The hardware complexity and costs must not be high.
The idea of modularity must be also reflected in the easy addition or removal of modules
(“plug & play”), facilitating an easy system set-up and maintenance.
It must allow a simple controller interface.
Error detection at the physical and the application layers are mandatory.
At the application layer, a protocol for off-line data transfer to/from the DCUs must be
available. This will allow the parameterisation and configuration of the DCUs. Data upload
from the DCUs will enable the off-line analysis of state variables saved at every 100 μs.
It should support Internet protocols and be compatible with other standard fieldbuses. It
should be also well established in the field of industrial automation.
Figure 2.12 Network connection in an industrial automation system
Control strategy 25
Over the last years RTE (Real-Time Industrial Ethernet) solutions for implementing industrial
fieldbuses gained popularity, so that nowadays there are many RTE-based protocols available [33].
The importance of Ethernet in the present industrial automation systems is shown in Fig. 2.12.
Even though at the beginning Ethernet was not considered an appropriate solution due to its lack
of real-time performance, the modification of this protocol at the upper levels of the OSI/ISO
model lead to new real-time capable, wire-based, protocols [34]. Taking in consideration the
requirements listed above and the actual trend for industrial network solutions nowadays, an
RTE-based protocol was used also in this work, namely EtherCAT. EtherCAT stands for Ethernet
for Control Automation Technology and is one of the most powerful industrial fieldbus available
now [35]. Due to the 100BASE-TX fast-Ethernet standard (specified in the IEEE 802.3), used at the
physical layer, high bit-rates of 100 Mbps are possible. It allows the connection of up to 65535
devices to the control network. At the fieldbus level there is a fix master-slave relation between
the CCU (EtherCAT master) and the DCUs (EtherCAT slaves). This protocol is completely hardware
implemented. The fieldbus interface of the master is a standard network card, so that a standard
PC can represent the CCU. The same network card is used to interact with both process and
control networks. No additional hardware is requested for the master. The fieldbus interface of
the DCU is implemented by EtherCAT ASICs.
The characteristics of EtherCAT and its role in the proposed system are presented in detail in
chapter 4.
27
3. Point-to-point connection between the distributed controllers
This chapter describes the operation of the data transfer between neighbouring DCUs, which was
outlined in the figures 2.9, 2.10 and 2.11. A more detailed description of the hardware, which is
used for this communication, is given later in the chapter “Experimental Set-up”.
3.1. SPI data transfer
SPI describes a synchronous and serial data bus. It was initially defined by Motorola, as a simple
and cost-effective method to transmit data between microcontrollers and peripherals like ADCs
and sensors. Along the same SPI-bus, the communication between two or more microcontrollers is
also possible. The data transfer is based on the master-slave principle. Data can be transmitted at
high rates (up to tens of Mbps) in full-duplex or half-duplex mode. The communication nodes on a
SPI-bus are typically connected over four logic signals:
A clock signal (CLK), which is generated by the master
A data signal (MOSI) from the master towards the slave(s)
A data signal (MISO) from a slave towards the master
A slave-select signal (SS), required if there is more than one slave-node on the bus.
SPI was not patented by Motorola (it is license-free) and is not standardized. The user can decide
upon the phase, polarity and rate of the clock signal, the size and shift operations of the incoming
and outgoing data, the number of slaves and slave-select signals and upon the communication
mode (half- or full-duplex).
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
SPI Unit(SPI Slave)
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
McBSP Unit(SPI Master)
CLK CLK
MOSI MOSI
MISO MISO
CLK
MOSI
MISO
CLK
MOSI
MISO
SS SS SS
DCU „n-1“ DCU „n“ DCU „n+1“
(dummy data)
(dummy data)
(dummy data)
(dummy data)
SPI = Serial Peripheral InterfaceMcBSP = Multichannel Buffered Serial Port
CLK = Bus ClockSS = Slave Select (active low)
MOSI = Master Output Slave InputMISO = Master Input Slave Output
Figure 3.1 SPI configuration for the distributed controllers
Point-to-point connection between the distributed controllers 28
The DSP TMS320F2812 of a DCU in this project contains two SPI-compatible units, which are used
to implement the connection with its adjacent DCUs. These units are: the SPI unit and the McBSP
unit, presented in the Fig. 3.1. McBSP is a synchronous data transmission protocol, which besides
the SPI-compatibility has also extended features, such as independent clocking for sending and
receiving data, multi-channel selection and frame synchronization [36].
The McBSP unit is here always working as a master and the SPI unit always as a slave. This was the
only configuration, for which the maximal speed of 12.5 Mbps, allowed by the used DSP-board,
could be reached. The slave-select signal was not used, because each slave unit is connected with
two master units. This active-low signal of the slave units was tied to ground, so that a slave unit
will always shift out its data if it receives a valid clock signal. A slave unit can exchange data only
with one master at a time. The selection of the active master unit is made by a communication
protocol with the help of four general-purpose I/O (GPIO) signals of the DSP. These signals assure
also that no short-circuits occur on the bus, as described later in the Figs. 3.4 and 3.5.
The master initialises the data transfer by sending the clock signal. The data are transferred as 16-
bit words. Both the master’s and the slave’s shift registers will shift their bits in and out according
to the phase and polarity of this clock. Here, the clock signal is inactive-low and has no phase
delay. The data is transmitted on the rising edge and received on the falling edge of the clock.
Fig. 3.2 shows the ideal case of the data transfer at a clock frequency of 12.5 MHz, where signal
delays due to the transmission medium have been neglected. The slave receives the rising clock
pulses generated by the master with no delay and after maximum 10 ns puts its data on the bus.
Along with the falling edge of the clock, both master and slave read the data. After 16 clock pulses
the master and the slave successfully exchange a word. This is a full-duplex communication mode.
The data transfer process of Fig. 3.2 corresponds rather to direct connections (i.e. single ended
transmission) over short PCB routes or cable lengths and cannot be used in this case. In our set-
up, due to reasons of EMC and longer cable distances, data must be transmitted differentially over
a one-meter long cable. We selected the differential signalling standard RS485 for the physical
layer of this communication. The propagation delay inserted by the RS485 transceivers and the
cable had to be considered for a full-duplex communication. A full-duplex communication at 12.5
MHz cannot be realised with the maximal propagation delays shown in the following table.
SS
CLK
MOSI
from Master
from Master
at SlaveCLK
MISO
Clock cycle 80ns
Bit 16 Bit 15 Bit 14 Bit 13 Bit 12 Bit 11
Bit 16 Bit 15 Bit 14 Bit 13 Bit 12 Bit 11
Writ
e
Rea
d
Writ
e
Rea
d
40ns
10ns
Figure 3.2 Ideal full-duplex communication process disregarding delays
Point-to-point connection between the distributed controllers 29
This is explained in Fig. 3.3, where the first bit read by the master has a false value and the second
bit read corresponds actually to the first bit send by the slave. All the bits read by the slave have
correct values.
As the full-duplex communication with real delays inserted by the RS485 transceivers doesn’t
work properly, in this project two simultaneous half-duplex transmissions were developed.
3.2. Communication protocol
The communication between any adjacent DCUs is initiated and stopped according to the relative
position of the vehicles. Fig. 3.4 shows the initial state where at the control stage only DCU “n” is
controlling the vehicle. The stator segment “n+1” is passive and at this moment there is no
communication process between the two DCUs. Concerning the communication, the SPI-
compatible units as well as the RS485 transmitters and receivers are outlined in the Fig. 3.4.
RS485 Transmitter Cable RS485 Receiver
10 ns (maximal delay value) 5 ns 15 ns (maximal delay value) Total maximal propagation delay: 30 ns
Interval Definition Value
T1 Clock cycle 80 ns T2 Delay time, SS low to CLK (master) high 40 ns T3 Total maximal propagation delay 30 ns T4 Minimal set-up time of a received bit before the falling edge of CLK (slave) 30 ns T5 Maximal delay time, CLK (slave) high to MISO valid 10 ns T6 Delay time, SS low to first bit read (master) 80 ns
Table 3.1 Timing specifications for a real full-duplex communication process
SS
CLK
MOSI
from Master
Master out
Slave in
CLK
MISO
Mas
ter
Writ
es
Bit 16 Bit 15 Bit 14 Bit 13 Bit 12 Bit 11
Slave datain
Slave dataout
Bit 16 Bit 15 Bit 14 Bit 13 Bit 12 Bit 11
MISOMaster data
in
SSat Slave
Bit 16 Bit 15 Bit 14 Bit 13 Bit 12
Mas
ter
Re
ads
Sla
veR
eads
Sla
veW
rites
Sla
veR
eads
Mas
ter
Rea
ds
T1
T2
False Bit Value
T3
T4
T5
T6
Figure 3.3 Full-duplex communication process considering real delays
Point-to-point connection between the distributed controllers 30
The communication protocol is based on four GPIO signals, two outputs and two inputs on the side
of each DCU. In order to avoid short circuits on the communication bus, the RS485 transmitters of
a DCU can be enabled/disabled only by the corresponding adjacent DCUs, as shown in Fig. 3.4.
Segment „n+1“Segment „n“
VSI x xx VSIx xx
DCU „n“ DCU „n+1“
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
TX
TXRX
RX
EN = 0
EN = 0
EN = 0
EN = 0 EN = 0
EN = 0
Co
ntr
ol
Po
we
r p
roce
ssin
gC
om
mu
nic
atio
n
DCU „n“ DCU „n+1“
RS485
RS485
(GPIO-1)
(GPIO-2)
(GPIO-3)
(GPIO-4)
Figure 3.4 Initial state of the point-to-point communication process
Segment „n+1“Segment „n“
VSI x xx VSIx xx
DCU „n“ DCU „n+1“
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
SPI Unit(SPI Slave)
McBSP Unit(SPI Master)
TX
TXRX
RX
EN = 0
EN = 0 EN = 0
EN = 0
DCU „n“ DCU „n+1“
RS485
RS485
EN = 1
EN = 1
Request
Acknowledge
12.5 Mbps
Communicationtriggering positionCon
trol
Po
we
r p
roce
ssin
gC
om
mu
nic
atio
n
12.5 Mbps
BOC
Figure 3.5 Full-duplex communication at 12.5 Mbps
Point-to-point connection between the distributed controllers 31
When the vehicle reaches the communication triggering position, marked in Fig. 3.5, DCU “n”
sends a communication request signal to DCU “n+1”, which enables at the same time the
transmitters of DCU “n+1” towards DCU “n”. DCU “n+1” must assess then its actual
communication state. If it is not already communicating with DCU “n+2”, as assumed in Fig. 3.5, it
has to acknowledge this communication request. Along with the acknowledge signal, the
transmitters of DCU “n” towards DCU “n+1” are enabled.
The entire point-to-point communication process is presented in the flowchart of Fig. 3.6. The
power stage of the modules is activated and deactivated also according to the relative position of
Initial state
CommunicationRequestreceived?
Communicationstatus
bits == 1x?
YES
Position BOCwas reached?
NO
Send Acknowledge signal &Communication status bits = 10 &
Wait two control cycle
Send Request signal &Communication status bits = 11 &
Wait two control cycle
CommunicationAcknowledge
received?
Communication processrunning
Position EOCwas reached?
Bring the vehicle in the middle of the stator segment &
Set communication error flag
Wait for the communication errorflag to be reset by the CCU
(over EtherCAT)
Error flag wasreset?
Check value ofthe communication
status bits
10 11
Stop sending data &Wait two control cycles &
Put Acknowledge signal low &Communication status bits = 00
Stop sending data &Wait two control cycles &Put Request signal low &
Communication status bits = 01
Communication status bits
00 communication inactive
01 communication inactive
10 communication running„slave“
11 communication running„master“
Vehicle positions:
BOC Begin of Communication
EOC End of Communication
NO
YES
YES
YES
NO
NO
NO
YES
YES
NO
Figure 3.6 Flowchart of the position dependent point-to-point communication process
Point-to-point connection between the distributed controllers 32
the vehicle, namely the power stage of module “n+1” is activated after the communication-
triggering position has been reached. Immediately after activation, the force producing current of
the module “n+1” is controlled to zero and there is no master-slave relationship between the
modules. DCU “n+1” becomes a slave controller only after it acknowledges the communication
request. Both DCUs exchange after this point simultaneously data (as shown in Figs. 2.9, 2.10 and
2.11) in both directions at the speed rate of 12.5 Mbps. This corresponds to a full-duplex
communication, but with no problems concerning propagation delays [37]. The communication
process ends when the triggering position “End of Communication (EOC)” is reached. The request
and acknowledge signals are set low, so that the RS485 transmitters of both communication parts
are disabled.
If DCU “n+1” communicates with DCU “n+2” and receives at this time a communication request
from DCU “n”, it will just ignore it. DCU “n” will not receive the acknowledge signal, so that after
two control cycles will stop immediately the vehicle. Afterwards it will set the collision flag inside
its state variable and will reposition the vehicle in the middle of the stator segment. The CCU must
then reset this error flag before sending new position reference values to DCU “n”. The CCU is
informed about the communication process by means of three bits inside the status variables of
all DCUs. One bit is a communication-error flag and the other two are reflecting the status of an
error-free communication (Fig. 3.6). During normal operation, collision situations should not
occur if the traffic planner of the CCU has been implemented correctly.
Remark: In the system, three levels of master-slave relations must be distinguished. At the level
of control strategy, the mastership moves according to the position of the vehicle and a new slave
controller is invoked when the vehicle transits to the next stator segment. For the SPI
communication, each DCU accommodates one SPI-master unit and one SPI-slave unit. For the
EtherCAT communication, which is introduced in the next chapter, an ordinary PC serves as
EtherCAT master and all servo-controllers are EtherCAT slaves.
The CCU starts via EtherCAT the control cycles in all DCUs simultaneously. This means that the
PWM-timers should run synchronized in all DCUs. But even though the electronic oscillators of
the processors inside the DCUs are of the same type, very small differences between their
frequencies still exists. This leads to a significant clock drift between them in few seconds.
Fig. 3.7 shows how are the data transmitted, received and read, when there is no clock drift
between two adjacent DCUs. The intervals 0 7 8 9, , ,T T T T are defined in Table 3.2. It is supposed that
the DCUs exchange the same amount of data (e.g. ten 16-bit variables) in each control cycle. In all
DCUs the SPI-received data is red shortly after a control cycle begins and data is transmitted at
the end of the cycle. Due to the synchronized transmit and read actions in both DCUs, the
transmitted data by one DCU in a control cycle are always received before being read in the next
control cycle by the other DCU. Fig. 3.8 shows the case when there is a clock drift between the
DCUs and not all transmitted data are received before being read.
Point-to-point connection between the distributed controllers 33
DCU “n+1” starts transmitting data shortly before DCU “n” reads its receive-FIFO. Only half of this
transmitted data will be available in the receive-FIFO at the time DCU “n” reads it. The other half
of the data will be read by DCU “n” in the next cycle. In the worst case the read data are not
consistent and there will be one cycle delay between transmitted and received data. A software
procedure keeps the consistency of the data inside the 16-level receive-FIFO of a DCU. No more
than ten variables are sent in a control cycle so that a FIFO overflow cannot occur.
ReadData
TransmitData
ReadData Transmit
Data
DataReceived
t
t
PWM-TimerValue
DCU„n“
DCU„n+1“
x x x x x x x x x x10 x16bits Variables
ReadData
ReadData
10 x16bits Variables
T7
x x x x x x x x x xx
ISR„n“
ISR„n+1“
T8 T9 T8
T0
Figure 3.7 Data transfer with synchronized PWM-timers
Interval Definition
T7 Between the read and transmit actions inside one control cycle (constant) T8 Required to transmit 10 x 16bits at 12.5 Mbps (constant) T9 Between the read action and the time when data were written in the FIFO (variable) T0 Control cycle (constant) T10 Between the transmit action in one DCU and the read action in the adjacent DCU
(variable)
Table 3.2 Timings related with data transfer and clock synchronization issues
ReadData
ReadData
ReadData
TransmitData
TransmitData
DataReceived
x x x x x x x x x xx--- - - - - - - -
- - - - -***** --- - - x x x x xx
TransmitData
ReadData
DataReceived
TransmitData
DCU„n“
DCU„n+1“
PWM-TimerValue
t
t
T8
10 x16bits Variables 10 x16bits Variables
ISR„n“
ISR„n+1“
T10 T8T10
Figure 3.8 Data transfer with non-synchronized PWM-timers
Point-to-point connection between the distributed controllers 34
Just one cycle delay in the communication process does not affect the control performance. The
current control loop runs independently in all DCUs and they exchange only current reference
values. The variation of the vehicle’s speed and consequently the induced EMF are assumed to be
zero within a current control cycle period (100 μs).
In principle, it would be possible to synchronize the PWM-timers of all DCUs by means of the
EtherCAT fieldbus. Any EtherCAT slave device (here the DCU) contains a special clock unit, which
can be used to synchronize, according to the IEEE 1588 standard, all the EtherCAT slaves to one
reference clock hold by one slave (typically the first slave after the master in a network segment).
But this would not be an advantage for this application. As a consequence of synchronization, the
IGBTs of two adjacent VSIs controlling a vehicle would always commutate approximately at the
same time instants. Due to the stray inductance of the common DC-link, synchronized switching
would increase the spikes in the DC-link voltage and the stress on the IGBTs.
3.3. Data transmission at the physical layer
In this application, where the electronic of all DCUs is tied to the negative DC-link rail, the safety
at the physical layer of the point-to-point data transmission is of concern, mainly due to common-
mode voltage problems. In this subchapter these problems are identified based on simulation and
experimental results and solutions are offered.
RS-485, also known as TIA/EIA-485, is a hardware standard, which describes a differential data
transmission over long cables in multipoint communication networks [38]. It assures a high level
of noise immunity and is therefore preferred in industrial environments. For a data channel a pair
of wires is required, one for the original and one for the inverted signal. Received and radiated
EMI are therefore (theoretically) eliminated, as these signals are equal but opposite on the two
wires and their fields will cancel each other. The maximal rate and distance of data transfer
depends though on the attenuation characteristics of the cable and the noise coupling of the
environment. A transmission rate of 12.5 Mbps over one meter long cable is possible.
Details of the hardware implementation are given in the chapter “Experimental set-up”. There,
Fig. 5.18 shows the implemented simplex network topology. For data and clock signals one
receiver is connected with two transmitters over the same pair of wires. Only one transmitter is
allowed to send at a time and the transmission is unidirectional. Termination resistances were
placed at the extreme ends of all pair of wires in order to match the characteristic impedance of
the wiring.
Two situations are of concern for proper signal transmission:
Idle-bus periods
Common-mode voltage problems
3.3.1. Idle-bus state
During the non-transition periods of the vehicle, the transmitters of data and clock signals are in
a high-impedance state, so that the voltage on the respective pair of wires is left floating. Due to
Point-to-point connection between the distributed controllers 35
the noisy environment, the corresponding receivers can be then falsely triggered and these
signals are erroneously interpreted as valid signals. This situation should theoretically not occur,
as the receivers are fail-safe. The input threshold value of their differential voltage is set between
–10 mV and –200 mV and not between ±200 mV as usual. Like this, due to the termination
resistances, their outputs should be logic-high during the bus-idle condition. Anyhow, the
bandwidth of this hysteresis voltage still doesn’t represent a reliable solution against high noise
levels. This problem is completely solved at the application layer by means of the communication
protocol. At this level, the bus is considered idle until the communication is acknowledged. During
this time the SPI-units are deactivated, meaning that the clock signals are ignored and no bit-
shifting actions take place inside their data shift registers. The transmitters for data and clock
signals are enabled when the communication is acknowledged, so that the bus is no longer idle.
The SPI- and McBSP-units are reset to initial configuration values and then activated. After a wait-
period of one control cycle (Fig. 3.6) the McBSP-units will write their clocks and data on the bus.
3.3.2. Common-mode voltage and multicommutation problems
In systems where a large number of inverters are connected to a common DC-link, high-voltage
transients can occur due to the multicommutation process of the IGBTs and the resulting stray
inductance of the DC-link connection.
The switching times of the IGBTs in the different phases of the distributed inverters are not
synchronized, never the less several IGBTs can switch simultaneously or nearly simultaneously.
This is known as multicommutation [39]. The multicommutation effect on an inverter system as
shown in Fig. 3.9 was analysed in [39], [40]. Here the voltage at the common stray inductance
,commonL is produced by the added /di dt values of the different commutating IGBTs.
Assuming the same /di dt for all IGBTs, the blocking voltage of an IGBT will be:
The value of the surge voltage ,CC surgeu must be kept bellow the maximal rating of the switching
devices. The first action to minimize this voltage is to keep ,commonL as low as possible. This is
Lσ,common
DC-
DC+
Lσ,ph1
CEL
Lσ,ph2 Lσ,ph3 Lσ,phn
Figure 3.9 Surge voltages for multiple inverter legs connected to a common DC-link [40]
, , , 1CC surge common ph D
di diu L n L U
dt dt (3.1)
Point-to-point connection between the distributed controllers 36
realized by implementing a compact bus system consisting of two copper sheets (DC+ and DC-),
separated by an insulation layer [41]. Of course it would be preferable to be able to limit or control
the value of /di dt , but the gate-drive circuit is not accessible in the commercial IPMs. Therefore,
the modules must be protected against high surge voltages by snubbing elements (usually
capacitors).
The simplified equivalent circuit of the proposed DC-link structure for this type of applications is
shown in Fig. 3.10 and the parameters are defined in Table 3.3. The DC-voltage supply should be
situated in the middle of the DC-link bus. A snubber capacitor is mounted very close to the DC-link
terminals of each IPM and the electrolytic capacitors are also distributed along with the inverters,
so that the current through the DC-link bus becomes more uniformly distributed [42].
Regarding the communication structure, the RS485 receivers have to be protected against voltage
transients on the common negative DC-link line between two adjacent modules. These transient
voltages are seen as common-mode voltages by the receivers and can have higher values than the
maximal specified ratings of the devices.
The value of the common mode voltage CMu (Fig. 3.11), seen by the RS485 receiver is:
Name Definition
Leff,ac Total stray inductance of the bus bar system between two modules Lσs Stray inductance of the PCB routes between the electrolytic and snubber capacitors LESR Equivalent series inductance of the electrolytic capacitors RESR Equivalent series resistance of the electrolytic capacitors CEL Value of the electrolytic capacitors LSN Equivalent series inductance of the snubber capacitors CSN Value of the snubber capacitors
LσIPM Stray inductance of the IPM
Table 3.3 Parameters of the circuit model of the common DC-link
DC-
DC+
LCU&
RUIPM IPM IPM
Module (n/2)-1 Module (n/2) Module (n/2)+1Leff,ac/2
Leff,ac/2 Leff,ac/2
Leff,ac/2 Leff,ac/2
Leff,ac/2Lσs
Lσs Lσs
Lσs
Lσs
Lσs
LESR LESR LESR
RESR RESR RESR
CEL CEL CEL
LSN LSN LSN
CSN CSN CSN
L σIP
ML σ
IPM
L σIP
ML σ
IPM
L σIP
ML σ
IPM
Figure 3.10 The circuit model of the common DC-link
2A B
CM line
u uu u
(3.2)
Point-to-point connection between the distributed controllers 37
The characteristics of the transient voltage lineu were first analysed with the help of a SIMPLORER
simulation model (Fig. 3.12), where both adjacent inverters were commutating simultaneously.
A real IGBT model was used in the simulation and a dead-time interval was considered for the gate
signals of an inverter-leg. Sinusoidal currents of different amplitudes and frequencies could be
generated on the passive RL-load configuration.
A
B
LocalGND „n“
LocalGND „n+1“
TXRX
uline
uCM uADC- uOUT
DCU „n“ DCU „n+1“
100Ω
50Ω
50Ω
uIN uB
Figure 3.11 Common-mode voltage issue for the RS485 receivers
Leff,ac/2
DC-
DC+
IPM
CSN
CEL
IPM
Lσs
LESR
RESR
LSN
L σIP
M
Reff,ac/2
200mΩ 35nH
75n
H17
5mΩ
330μF
5nH
1μF
15nH
5nH
5nH
560V
uline
iline
iC1 iC2
35nH200mΩ
Rload
Lload
Leff,ac/2Reff,ac/2 Leff,ac/2Reff,ac/2
Leff,ac/2Reff,ac/2
Lσs
Lσs
Lσs
LESR
RESR
CEL
LSN
CSN
L σIP
M
L σIP
ML σ
IPM
Rloa d
Lload
Figure 3.12 Simulation model for the analysis of the transient voltages on the negative DC-bar
1 mm
Insulator 0.5 mm
Copper barsDC+
DC-
30 mm
5 mm
2000 mm
600 mm
Bars fixingScrew
Top view
Side view DC+
Figure 3.13 Bi-plate bus system as common DC-link bus
Point-to-point connection between the distributed controllers 38
a1) & b1) linei : Current through the DC-link bars between the two inverters
a2) & b2) 1Ci : Current through the electrolytic capacitor of the first inverter
a3) & b3) 2Ci : Current through the electrolytic capacitor of the second inverter
a4) & b4) lineu : Transient voltage along the common negative DC-link between the inverters
The equivalent circuits of the electrolytic and the snubber capacitors were considered in the
simulation. The values of their series resistances and inductances are calculated by means of their
impedance values at the respective resonance frequencies. The impedance of the planar bus
system consists of resistance and inductance. The inductance of each bus conductor is the sum of
the self-inductance barL and the mutual inductance mtL . The bus system is a bi-plate bus system, as
shown in Fig. 3.13, where the currents in both bars have opposite directions. For a good inductive
Figure 3.14 Simulation results for the considered simulation model
Point-to-point connection between the distributed controllers 39
coupling between the bars, the distance between them has to be small. The total inductance of the
bus system is:
The values 1, 2,,bar ac bar acL L represent the self-inductances at high frequency of the rails DC+ and DC-
respectively. Typically, the frequencies of the transient voltages on these lines reach values up to
several MHz. The skin- and proximity-effects play therefore an important role in determining the
resistance and inductance values of the planar bars, as the inductance ,eff acL decreases at high
frequencies [41].
The equations defining the effective inductance of the implemented bi-plate bus system (Fig. 3.13)
at low and high frequencies as well as the resistance value at high frequency [41], [43], are
presented in the appendix. The bus bar length between two adjacent servo-controllers is 0.6 m.
Due to the connection elements of the servo-controllers to the bus bars, the value of ,eff acL for the
mentioned length was approximated for the simulation at 70 nH.
The simulation results presented in Fig 3.14 show that the value of the transient voltage lineu is
below the maximal allowed value of the common-mode bus input range, ±15 V, for most RS485
devices. The parameters of the DC-link components are shown in Fig. 3.12. The currents on the
DC-link bus are shared between both DC-link capacitors. For the entire system these currents will
be shared between the capacitors of both active and inactive inverters. The frequency of the
transient voltage lineu is about 5 MHz.
c) lineu : Transient voltage for , / 2 10eff acL nH and 15SNL nH
d) lineu : Transient voltage for , / 2 35eff acL nH and 30SNL nH
General conclusions of the simulation results:
The stray inductances ,S IPML L have a very small influence upon the value of lineu but a
big influence upon the surge voltages over the IGBTs.
The inductances of the snubber capacitors and of the bus bars have the biggest influence
upon the value of lineu (Fig. 3.15) and should be kept as small as possible.
, 1, 2, ,2eff ac bar ac bar ac mt acL L L L (3.3)
Figure 3.15 Effects of the DC-link parameters upon the transient voltage lineU (simulation)
Point-to-point connection between the distributed controllers 40
The values of the resistances of the DC-link components are not relevant for the amplitude
of the voltage lineu . They only contribute to the damping time of the transient process.
The simulation results help to identify those circuit elements, which are most relevant for the
generation, suppression and the characteristics of the voltage transients, but cannot substitute
the measurement results. With the simulation model is not possible to consider disturbances, as
parasitic capacitances and interference effects, which exist in a real set-up. Also an exact
determination of the parameters of all circuit elements and the exact switching characteristics of
the IPM cannot be obtained.
Anyhow, the experimental measurement from Fig. 3.16 shows similar results as in the simulation
for the same control conditions. The damping time of the oscillations is larger but their
amplitudes are not of concern. Using a tri-layer planar bus system would further reduce the
effective impedance of the bus [41], so that lower transient voltages could be achieved.
In systems where the voltage transients are still of concern, additional protection measurements
must be used. A first possible solution to protect the RS485 devices against high common-mode
voltages, which luckily do not exist in the proposed application, is to isolate the ground of these
devices from the host’s device ground by using optical or digital isolators (Fig. 3.17). This method
Figure 3.16 Experimental measurement of the transient voltage lineu
A
B
LocalGND „n“
LocalGND „n+1“
TXRX
DC-
DCU „n“ DCU „n+1“
100Ω
50Ω
50Ω
Isolated GND
Isolated supply lineLocalsupply
line
Iso
lato
r
Localsupply
line
Isol
ator
uIN
uCM
uline
uA uB
uOUT
Figure 3.17 Galvanic isolation of the RS485 devices (not necessary in the proposed system)
Point-to-point connection between the distributed controllers 41
requires, beside the isolators, also an isolated power supply between any two adjacent modules. As
one of the aims of this work is to keep the implementation costs for large systems low, isolation
should be avoided as far as possible.
Another, less-expensive solution is to use bi-directional transient voltage suppression (TVS)
diodes. These are clamping devices, which suppress the overvoltages above their breakdown
voltage and can withstand high power peaks (several hundreds of Watts). They are connected as
shown in Fig. 3.18 and are selected to conduct at voltages in the proximity of the recommended
operating conditions of the RS485 devices. The disadvantage in this case is that these diodes add
an extra capacitance to the communication lines, which deteriorates the transmitted signals at
high bit-rates.
The receiver used in the experimental set-up (SN75LBC175A) has a built-in circuit protection
against overvoltages, realised with standard Zener diodes, which have a very small capacitance,
which doesn’t affect the transmission speed. It can support short voltage transients in the range
of ±30 V but with a maximum power peak of approximately 9 W. This is sufficient for the proposed
application and is the most cost effective solution.
A
B
LocalGND „n“
LocalGND „n+1“
TXRX
DC-
DCU „n“ DCU „n+1“
100Ω
50Ω
50Ω
TSVDiodes
uIN uCMuA uB uOUT
uline
Figure 3.18 Transient voltage protection with TSV diodes
43
4. Industrial fieldbus solution for motion coordination and monitoring
To connect the CCU with all DCUs an industrial fieldbus is selected. During development,
commissioning and test of a large system, control programs must be frequently modified and then
downloaded to a multitude of DCUs. Also upload actions of process data, recorded within the
DCUs, have to be possible. Up- and downloads are off-line file transfers initiated on demand. While
the system of linear drives is in normal operation, data for motion coordination and monitoring
have to be exchanged cyclically between CCU and the DCUs inside few milliseconds.
4.1. Overview of the main real time Ethernet (RTE)-based fieldbuses
The actual trend in distributed automation systems is to use Ethernet as the single networking
technology, as shown in Fig. 2.12. The main characteristics of Ethernet, which lead to this trend,
are [44], [45]:
The high data transmission rates at the physical layer, which assure a large bandwidth
compared with other fieldbuses
At the network and the transport layer of the OSI model, the IP and the TCP/UDP protocols
allow an easy integration in the Internet and the use of application layer protocols such as
FTP (up- and download of general data to the field devices) or HTTP (web servers for device
engineering).
The implementation costs are low as the technology is wide spread and well specified
Because Ethernet alone is not real-time capable due to its nondeterministic arbitration
mechanism, several techniques (e.g. modified CSMA, token passing, TDMA, master-slave approach
and switched Ethernet), meant to transform Ethernet in a true real-time communication tool,
have been developed over the years [33], [46], [47]. The RTE protocols can be allocated to one of
the three possible architectures shown in Fig. 4.1. The first architecture uses Ethernet and the
TCP/IP protocols without any modification. Only at the application layer a specific protocol is
defined. The second architecture uses the standard Ethernet hardware but bypasses the TCP/IP
stack, so that a specific protocol type is specified in the Ethernet frame. The third architecture
also bypasses the TCP/IP stack but brings also modifications in the hardware of the Ethernet. The
scheduling process of switched Ethernet is modified and the switching functionality is integrated
in the field devices. These types of protocols achieve the highest performance (e.g. have data
Industrial fieldbus solution for motion coordination and monitoring 44
update cycles smaller than one millisecond) and are therefore preferred in distributed servo-drive
applications. The protocols belonging to the second and third architecture are also able to handle
standard IP frames, either on the same channel as the real-time data (tunnelling) or on a separate
channel.
There are many RTE protocols, which can fulfil the task of motion coordination and monitoring
for this application, regarding the necessary data update cycle of few milliseconds. Many of them
have been created for specific applications and just few were widely adopted and internationally
standardized, like e.g. Ethernet Powerlink, EthernetIP, PROFINET, EtherCAT and SERCOS III.
NonReal-TimeProtocol
Real-TimeProtocol
TCP/UDPIP
Ethernet
NonReal-TimeProtocol
Real-TimeProtocol
TCP/UDPIP
Ethernet Modified Ethernet
NonReal-TimeProtocol
Real-TimeProtocol
TCP/UDPIP
e.g. ModBus, EthernetIP e.g. Ethernet Powerlink, EPA
e.g. EtherCAT, SERCOS III PROFINET IO
1) On top of TCP/IP 2) On top of Ethernet 3) On top of modified Ethernet
Figure 4.1 RTE architectures [46]
Fieldbus characteristics EtherCAT SERCOS III PROFINET IRT
Principle of operation
Master/Slave
Master/Slave
Time synchronization of the device internal
switches
Special Ethernet controller Required only for the
slave devices Required for both master and slave
devices
Required
Basic network topology
Line, ring, star or tree structures are
supported
Line or ring
Line or ring
Cable redundancy
Yes
Yes
Yes
Overall Performance [48]
Very good
Very good
Good
User organization
ETG (EtherCAT Technology Group)
SI (SERCOS International)
PNO (PROFIBUS Nutzerorganisation)
Number of suppliers, products variety and
organization members
Big
Medium
Medium
Table 4.1 Overview of the main characteristics for three RTE-based fieldbuses
Industrial fieldbus solution for motion coordination and monitoring 45
The utilization of an RTE-based fieldbus for motion coordination and monitoring in this case
means that the DCUs must incorporate a hardware interface for this fieldbus. An ASIC, an FPGA, a
multi-protocol Ethernet controller or a standard Ethernet controller can represent this hardware
interface. Due to the high number of DCUs in our application, the hardware costs and complexity
is one of the main criteria in selecting the fieldbus.
The costs for the special Ethernet controllers are quite similar for the three protocols of Table 4.1,
because the respective protocol stack can be implemented in an e.g. FPGA. EtherCAT has the
advantage that its master device requires only a standard and no special Ethernet controller.
EtherCAT offers also the most flexible network topology and has the largest variety of products,
number of suppliers and organization members [35]. For smaller, non-scalable system topologies
(small number of vehicles) where the communication between the adjacent DCUs would have
been possible over an RTE-based fieldbus alone (no extra SPI-based fieldbus required), then
SERCOS III would have been the better solution due to its superior performance regarding the
cross-communication procedure between the slave devices [34].
4.2. Data transfer modes over EtherCAT
EtherCAT is a master/slave-based protocol where the master always initiates the data
transmission inside a communication cycle. The master can use the same Ethernet frame to send
data to, or gather data from the slave devices. Inside the slave devices the incoming frames are
hardware processed “on the fly”, meaning that data is added to or extracted from the frame while
this passes through the device. The delay-time due to data processing of a slave is very short
(below 100 ns) and constant, as it doesn’t depend on the size of the frame. When the formed frame
reaches the last slave of the network, it is automatically sent back to the master. The slave devices
of an EtherCAT network are designated to one or more network segments. A segment consists of
one or several slaves. From the standard Ethernet point of view an EtherCAT segment is a single
Ethernet device. There are two configuration modes of an EtherCAT network:
Direct mode, where a network segment is designated to a master device
Open mode, where several network segments are accessed separately over a switch by one
or several masters
The slave devices of a network segment can be arranged in a line, ring, tree or star configuration.
All configurations can be seen as a physical ring due to the full-duplex data transmission and to
the fact that the slave devices can transfer data also in the reverse direction [49].
The Ethernet frames used by EtherCAT are identified with the number 0x88A4 in the header of the
frame. The data in the frame is split in specific EtherCAT datagrams as shown in Fig. 4.2. Any
datagram can be designated to a particular slave but an important feature of EtherCAT is that
many slaves can be handled by a single datagram using a logical addressing scheme. The logical
address space of 4GB is shared among all slave devices of a network segment. Logical addresses
are converted inside a slave device into local physical addresses by means of the Fieldbus Memory
Industrial fieldbus solution for motion coordination and monitoring 46
Management Units (FMMUs). The FMMUs allow therefore the logical addressing for data segments
that span over several slave devices. This reduces the total overhead.
Until now it was shown how can the master access the memory of the slave devices. This memory
is also accessed by the DSPs of the DCUs. Hardware details are shown in the Fig. 5.19 of the
chapter “Experimental set-up ”. A memory access management is therefore necessary. This is
realized by the SyncManager units of the slave devices, which are using memory buffers of
predefined lengths to assure a consistent and secure data exchange between the EtherCAT master
and the DCUs. There are two communication types in an EtherCAT network:
Reliable communication (mailbox data transfer), where the master and the DCU accessing
the memory of the ESC must agree upon the write/read access (data cannot be
written/read at/from one address by one unit until it was before read/written by the other
unit).
Non-reliable communication, where both the master and the DCU can access the memory
of the ESC at any time for both read and write actions.
The reliable communication is based on a specific protocol at the application layer, which must be
also implemented in the DCUs. Here the FoE (File access over EtherCAT) protocol was used for
both update and parameterisation (download) of the control algorithm as well as for general data
upload. This protocol is similar to the well-known FTP protocol. The general data, representing
state variables of the control algorithm, were saved in every control cycle in the memory of the
DCUs. The master then uploaded them off-line, where they were analysed with MATLAB to prove
the quality of the control. For this communication the master used physical addressing and
exchanged data only with one DCU at a time.
The non-reliable communication is used for cyclical exchange of process data. The term non-
reliable is associated with the fact that old data will be dropped if the above-mentioned memory
Figure 4.2 Structure of the Ethernet frame used by EtherCAT [49]
Industrial fieldbus solution for motion coordination and monitoring 47
buffers are written faster than being read. This communication serves for motion coordination
and monitoring. Based on an internal traffic planner, the master generates cyclically the new
position reference values for the vehicles. These values are embedded in an Ethernet frame as
shown in Fig. 4.3.
As this frame passes through, the DCU will notice that its logical address (predefined in its FMMU)
is inside the logical memory area specified in the header of the first datagram (logical start
address and length) and extracts the appropriate position reference value from it. Similarly
occurs with the second datagram, when the DCUs insert their status values in the frame. After
being processed by all DCUs, the frame returns to the master for evaluation. The master checks
first the valid processing of a datagram by comparing the WKC (Working Counter) with the
expected value. This counter is incremented by the DCUs after a successful write or read action.
The frame ends always with the FCS (Frame Check Sequence). This is recalculated inside each DCU
and if an error is detected, the master will be flagged by this DCU. A fault is like this precisely
located in the system [49].
EthernetHeader
Datagram1Header
X10 Xm X2Xm WKC S1 S2 SnS3 WKCDatagram2Header
FCS
LogicalWrite
LogicalRead
X10 S1 Xm S2 X2 Sn
Ether Type0x88A4
MotionCoordination
Algorithm
Monitoring
EtherCAT Master 4GB Logical Address Space
Start Address
Start Address
x
Segment 1 Segment 2
x x x x x x x x x x x
Segment 3 Segment n
DCU 1 DCU 2 DCU 3 DCU n
Vehicle 10 Vehicle m Vehicle 2
Position X10Position XmPosition Xm
Position X2Position reference value for Vehicle 2written at the address designated forSegment n
Status S1Status S2Status S3
Status Sn Status value of the last segment
Xm S3
Linear direct drive withn stator segments and
m vehicles
Figure 4.3 Cyclical data transfer for motion coordination and monitoring
Industrial fieldbus solution for motion coordination and monitoring 48
4.3. The EtherCAT master application
The Windows Control and Automation Technology (TwinCAT) is a software application running
under Windows NT, which implements the functionality of the EtherCAT master [50]. It gives the
user the possibility to define special tasks, which have different priority levels and are executed
cyclically at specified time intervals. The lowest cycle time is 50 μs. The user cannot write control
code or any code lines inside those tasks but can only define variables. In each communication
cycle the defined input-type and/or output-type variables of a task can be exchanged with the
process data of any EtherCAT slave device. An input variable of a task can be also associated with
an output variable of the same task, so that process data communication between two EtherCAT
slaves is possible. For the experimental set-up an output variable of a task, representing the
DCU 1
DCU 2
DCU 3
DCU 4
DCU 1
DCU 2
DCU 3
DCU 4
PDO 1
PDO 2
PDO 3
PDO 4
PDI 1
PDI 2
TWINCAT SOFTWARE TASK
IN 1
IN 2
IN 3
IN 4 OUT 4
OUT 1
Inputs Outputs
PDO = Process Data OutputPDI = Process Data Input
OUT 3
OUT 2
PDI 3
PDI 4
Figure 4.4 Analysis of TwinCAT’s real-time functionality
Industrial fieldbus solution for motion coordination and monitoring 49
position reference value of the vehicle, was associated with the input process data of all EtherCAT
slaves. Output process data, representing the status of the distributed servo-controllers, were in
turn associated with four input variables of the same task. Access to the variables of a task can be
gained from Windows applications like Visual Basic or Visual C++ via a specific DLL. Here a Visual
Basic application was created, which generates the new position reference values for the vehicle
and monitors the system at 10 ms intervals.
The real-time capability of TwinCAT was analysed as shown in Fig. 4.4. A program in the first DCU
toggled cyclically the value of an output process data and output this action on a GPIO. The
program in the other three DCUs read continuously (pooling) the memory of the local ESC and
output a physical signal according with the value of one input process data. The output process
data of the first DCU was connected with an input variable of a software task running at 50 μs. The
input process data of the other three DCUs were connected with three output variables of the
same software task. The software task was therefore responsible for the real-time data transfer
between the DCUs. The minimal data transfer delay is about 56 μs (one cycle) and the maximal
delay is 106 μs (two cycles). This is due to the fact that the cyclical writing of the process output
data of the first DCU is not synchronized with the Ethernet frames sent by the software task. The 6
μs difference represents the latency of the system.
The communication relationship at a certain moment between the master application and the
DCUs is defined by means of the EtherCAT state-machine, shown in Fig. 4.5. The master requests a
state transition through a special register of the ESC. The transition requires specific
initialisations of the units (e.g. FMMUs and SyncManager) and the memory registers of the ESC
from both DSP and master side. The EtherCAT state-machine must be therefore implemented in
the DSP-Flash memory.
INITIAL (INIT)
PRE-OPERATIONAL(PO)
BOOTSTRAP (BS)
SAFE-OPERATIONAL(SO)
OPERATIONAL(OP)
State Description
INIT The master initialises the network Each DCU receives a fixed address
BS Reliable communication with the DCUs using the FoE protocol
PO Reliable communication using other protocols (e.g. CANopen)
SO
System configuration for cyclic data transfer. Evaluation of process input data only
OP
The software task is started. Cyclic process data exchange between the master and the DCUs
Figure 4.5 Description of the EtherCAT state-machine
Industrial fieldbus solution for motion coordination and monitoring 50
4.4. Software configuration for the embedded servo-controllers
The DCUs must work as embedded systems, where the entire firmware is saved in the internal
Flash memory of the DSPs [51]. Initially, in order to write the code sections in the DSP-Flash, the
JTAG interface of the DSP-board is used. This implies a non-isolated connection between the CCU
and the DCUs. The DSP-Flash has two partitions. The first partition contains code sections, which
are not subject to changes and the second contains modifiable code sections (Table 4.2).
Fix Code Sections Changable Code Sections
1). Initialisation of system control registers and of the Flash unit
1). Initialisation of variables used for the control algorithm
2). Initialisation of the used peripheral units (ADC, EVA, SPI, McBSP, XINTF)
2). The motor control algorithm (the field oriented control)
3). Transfer of functions from Flash to RAM. Running code from Flash reduces the processing speed at 80%
3). Functions defining the serial point-to-point communication
4). Functions to Erase, Program, Verify the Flash 5). The EtherCAT state-machine 6). The FoE protocol
Tabelle 4.2 Main code sections of an embedded servo-controller
Reset DSP
Write code. Compile and link files
Execute and debug code
Use CCS and the JTAGinterface to write the code
in the DSP-Flash
Begin execution at entrypoint in RAM. The entrypoint is set by jumpers.
Reset DSP
Run the ProgramState-Machine
Begin execution at entrypoint in Flash. The entrypoint is set by jumpers.
Use TwinCAT as Flashprogrammer
Use hex2000 utility to convert the output file
COFF
Write code. Compile and link files using
CCS
Load Code in Flash Run Code from Flash
Modify Code in Flash
SystemInitialization
BS=?
Bootstrap
ChangeCode
?Run FoEprotocol
Reset DSP(use Watchdog)
OP=?
Operational
PO=?
Pre-Operational
Bootstrap
Run FoEprotocol
INIT=?
Reset DSP(use Watchdog)
BS=?Download
Data?
Yes
No
Yes
No
No No
No
No
No
Yes
Yes
Yes
Yes
Yes
Figure 4.6 The state-machine of the embedded servo-controllers
Industrial fieldbus solution for motion coordination and monitoring 51
Once the EtherCAT state-machine and the FoE protocol have been downloaded in the DSP-Flash, it
is possible to update the changeable code sections or to download general data by using TwinCAT
as Flash-programmer or Flash-reader respectively. The down- and uploading process over the
physical layer of Ethernet has the advantage that the communication partners are electrically
isolated.
In order to update some code sections of a DCU, the program is written and compiled with the
integrated development environment CCS (Code Composer Studio), installed on the CCU. From the
generated COFF (Common Object File Format) output file, the modified code sections are extracted
and saved in ASCII format with the help of the hex2000 conversion utility [52]. They are further
passed to TwinCAT, which will address the desired DCU by its physical address and will initiate the
reliable communication with it over the FoE protocol. After the data transfer is finished, the DSP
is reset with the help the Watchdog and the new code can be used without having to power-down
the controller.
Fig. 4.6 shows the main operations performed on the DSP-Flash and the state-machine for a servo-
controller. This is based on the EtherCAT states of the DCU. The interrupt routine for the control
algorithm is started always in the operational state. From the Bootstrap state is possible to update
the code sections, which requires a reset at the end, or simply to download general data for
analysis without resetting the device.
Industrial fieldbus solution for motion coordination and monitoring 52
5. Experimental set-up
In Fig. 1.3 the general configuration of a linear drive system was shown. This chapter presents the
implementation of all units of the drive system, but focuses more on the Information Processing
Unit (IPU) and the Power Processing Unit (PPU). The developed system architecture was tested in
a small-scale experimental set-up, consisting of four identical stator segments and one vehicle.
Each segment has a dedicated servo-controller. The block diagram representation of the
experimental set-up is shown in Fig. 5.1 and the photo of the set-up is shown in Fig. 5.2.
5.1. The Linear Motor Unit
Four stator segments of type LSE10G from Baumüller are building the track in a single-cam
arrangement (Fig. 5.3). The stator windings are distributed in the slots of a laminated iron core as
shown in Fig. 2.1. The skewed PM is bounded on the surface of a massive back iron (Fig. 5.4). The
mechanical construction of the vehicle–guideway system is shown in Fig. 5.5.
DCUbased onthe DSP
TMS320F2812
SensorInterface
SPIInterface
EtherCATSlave
Interface
DCUbased onthe DSP
TMS320F2812
SensorInterface
SPIInterface
EtherCATSlave
Interface
DCUbased onthe DSP
TMS320F2812
SensorInterface
SPIInterface
EtherCATSlave
Interface
based onthe IPMPS22056
VSIbased onthe IPMPS22056
VSIbased onthe IPMPS22056
VSIbased onthe IPMPS22056
VSIDC+
DC-
-280
V)
DCUbased onthe DSP
TMS320F2812
SensorInterface
SPIInterface
EtherCATSlave
Interface
x xx x xx x xx x xx
RS485
RS485
Ethernet
From theSensorModule
Segment 1 Segment 2 Segment 3 Segment 4
Ele
ctron
ic's Gro
un
d
Position Sensor Scale
CCUbased on
a standardPC
OperatorInterface
EtherCATMaster
Application
EtherCATMaster
Interface
4 KV LANMagnetics
Ethern et
GND
Mains3 ~
400V,50Hz
Line Connection
andRectifier
x5x3x3x3x3
PE
ServoController 1
ServoController 2
ServoController 3
ServoController 4
LN
x3 x3 x3 x3
Figure 5.1 Block diagram of the experimental set-up
Industrial fieldbus solution for motion coordination and monitoring 53
The vehicle maintains a fix vertical position along the
guideway due to the four-wheel system on the guide rail,
so that there is a constant air gap of 1 mm between the
PM and the stator segments.
The pole pitch of the motor is 36 mm. The PM has 4 poles
and a stator has 14 poles. Due to the even number of
poles of a stator and the alignment of its windings, there
VehicleServo-
Controller(DCU+VSI)
CCU
DC-link bus4 KV LANMagnetics
Line Connection& Rectifier
MotorCable
Figure 5.2 Photo of the experimental set-up
SE
GM
EN
T 4
SE
GM
EN
T 3
SE
GM
. 2
Figure 5.3 The linear track
Figure 5.4 The vehicle
Industrial fieldbus solution for motion coordination and monitoring 54
is no phase-shift of the electrical angle between two adjacent stators.
An odd number of stator poles would have assumed a phase-shift of 180°. The flux linkage of a
phase winding at the end of a stator will continue in the same phase of the following stator. Even
180m
m
135
mm
154mm
140
mm
90mm
74
mm
135m
m
45mm
19mm
32mm
45m
m
19m
m
90mm
65mm
1m
m
16mm
Fundament of the track
Magnet & Back Iron
Stator Segment
Fundament of the Stator
Guide rail
32m
m
42m
m
VehicleTop View
Front View
12m
m
Figure 5.5 Mechanical construction of the vehicle and of the guideway
Industrial fieldbus solution for motion coordination and monitoring 55
though there is no gap between the
stators, there is a gap between the end-
windings of two adjacent stators due
the thickness of the stator enclosure. In
this area the value of the reluctance
will increase, leading to a decrease of
the magnetomotive force for the same
flux density.
Since the PM covers only 28.5% of the
stator’s surface, it is expected to have a
low induced EMF-value. The force
parameters in Table 5.1 are specified
for the case when the length of the PM
equals the stator length.
5.2. The Power Processing Unit
The PPU contains the following functional blocks:
Line Connection Unit (LCU)
Rectifier Unit (RU)
Voltage Source Inverter (VSI)
5.2.1. The Line Connection Unit
The LCU contains the main switch, a line circuit breaker, line-fuses and an EMC-filter, as shown in
Fig. 5.6. Over the main switch the electronic of the servo-controllers is supplied first. The line
circuit breaker is closed only after a safe power-up of the electronic. Variable frequency drives are
a source of electromagnetic interference. This electrical noise is generated by the fast switching
voltages at the motor and supply cables, which have coupling capacitances to earth. An EMC-filter
is therefore necessary to assure that this high-frequency electrical noise, in the range of a few kHz
up to tens of MHz, doesn’t affect the mains. At the same time it assures that electrical noise in the
mains, generated by other equipments, doesn’t affect the drive system. The used EMC-filter is
rated at 25 A.
5.2.2. The Rectifier Unit
The RU contains a standard 3-phase diode bridge rectifier rated at 1200 V/25 A, a brake chopper
and a pre-charge resistance connected over a circuit breaker between the EMC-filter and the
Maximal thrust force ,maxthF = 1270 N
Nominal thrust force ,th nomF = 210 N
Maximal current maxI = 13.9 A
Nominal current nomI = 1.9 A
Force constant Fk = 110 N/A
Phase resistance SR = 2.4 Ω
Phase inductance SL = 10.5 mH
Pole pitch length = 36 mm Stator segment length
ml = 504 mm PM length
pl = 144 mm
Maximal speed maxv = 4.5 m/s
Nominal speed nomv = 2 m/s
Mover’s weight vM = 6.5 Kg
DC-link voltage DU = 560 V
Cooling type Natural cooling
Table 5.1 Parameters of the linear motor
Industrial fieldbus solution for motion coordination and monitoring 56
diode bridge (Fig. 5.6). The commutation notches of the diode bridge will insert current harmonics
in the mains. This will affect the ideal sine-wave voltage characteristic at the input of the rectifier.
In industrial plants where many drives are connected to the mains, the generated total harmonic
distortion is of concern and have to be diminished by inserting line chokes at the LCU level.
5.2.3. The Voltage Source Inverter
The VSI builds the modular servo-controller together with the DCU (Fig. 5.1). There was no
commercial solution to completely suit the requirements of this application, so that the modular
servo-controller was designed and implemented in the laboratories of the institute.
The VSI contains the following components:
6 IGBTs and the associated free-wheeling diodes
Gate driver circuit
Current measurement and overcurrent protection circuit
DC-link capacitors
DCU-interface circuit
Electronic power supply circuit
The power range of the inverters for this type of application lies between 1 KW and 10 KW. For a
DC-link voltage of 560 V, the rated currents of the inverters are in the range of 10 A up to 70 A.
EMC-filter
L1 L2 L3NPE
K1
Rc
VSI1 VSI2 VSI3 VSI4
560V (DC)
BrakeChopper
N
230V (AC)
Main Switch
K2
Electronic Supply
DC-link Supply
LCU
RU
DC +
DC -
DC +
DC -
F25A
F10A
100Ω
F25A
F25A
400μF
400μF1200V
25A
L1'
L2'
L3'
Figure 5.6 Configuration of the PPU
Industrial fieldbus solution for motion coordination and monitoring 57
Intelligent power modules (IPMs) represent a cost effective and highly integrated solution for the
power stage of highly scalable drive applications [53]. The IPMs use the transfer molding
technology for insulation and housing and were designed initially for home appliances (600 V-
IGBTs). About five years back, transfer molded IPMs in dual in-line packages (DIP) with 1200 V-
IGBTs and currents up to 35 A became available for general industrial application use.
Here the 1200 V/25 A DIP-IPM, PS22056, from
Mitsubishi was used. This device contains within
the same package the power devices and control
ICs for gate drive and protection [54]. The
molded structure is shown in Fig. 5.7. The
internal circuit and the pin-configuration are
shown in Fig. 5.8.
Each IGBT from the high side is controlled by the
HVIC (High Voltage Integrated Circuit). This
integrates high voltage level shifters, which
make a direct, non-isolated, connection to the
DCU possible. The IGBTs from the low side are
controlled by the LVIC (Low Voltage Integrated Circuit). If the HVIC is supplied over a bootstrap
circuit, a single 15 V source is sufficient to supply the entire gate drive circuit.
The integrated protection circuit protects the device in case of overcurrent and under-voltage
situations. Both the HVIC and the LVIC have an integrated under-voltage protection circuit (UVP).
Heat Sink
CreepageDistance
Figure 5.7 1200V DIP-IPM [54]
SignalCond. &
LevelShifting
GateDrive
&UVP HVIC
GateDrive
SignalCond.
ProtectionCircuit
FaultLogic &
UVP
PPS22056
GateDrive
&UVP
GateDrive
&UVP
HVIC
HVIC
SignalCond. &
LevelShifting
SignalCond. &
LevelShifting
LVIC
U
V
W
NU
NV
NW
VUFB
VUFS
VP1
UP
VVFB
VVFS
VP1
VP
VWFS
VWFB
VP1
WP
VPC
VN1
UN
VN
WN
FOCFO
CIN
VNC
Positive DC-link terminal
U-phase motor terminal
V-phase motor terminal
W-phase motor terminal
U-phase negative DC-link terminal
W-phase negative DC-link terminal
VUFS, VVFS, VWFSHigh-side drive supply GND terminals
VUFB, VVFB, VWFB
UP, VP, WPHigh-side control input terminals
UN, VN, WNLow-side control input terminals
High-side drive supply positiveterminals
VP1High-side control supply positive
terminals
VN1Low-side control supply positive
terminals
VPCHigh-side control supply GND
terminals
VNCLow-side control supply GND terminals
FOFault signal output terminal
CFOFault pulse time-setting terminal
CINOvercurrent detecting terminal
V-phase negative DC-link terminal
Figure 5.8 IPM PS22056: Internal circuit and pin-configuration
Industrial fieldbus solution for motion coordination and monitoring 58
If the voltage drops below the 12.5 V threshold value, the IGBTs will be automatically turned off.
The LVIC will also generate a fault output (FO) signal. If an under-voltage situation occurs on the
HVIC-side, the high side IGBTs will be turned off but no fault signal will be generated.
An external circuit is responsible for the detection of an over-current. The INC terminal of the
IPM is connected with the output of this circuit. If the voltage at this pin exceeds a certain
threshold value (typically 0.5 V), the protection circuit of the IPM will recognize the over-current
situation and will turn off all IGBTs. At the same time the FO signal is generated. The pulse width
of this signal can be set over an external capacitor between the terminals FOC and NCV .
The characteristics of the IPM (e.g. maximal ratings, static and switching characteristics) are
presented in the appendix.
Regarding the heat sink for the IPM, it must be mentioned that this had to be connected to the
negative rail of the DC-link (ground of the electronic). There are parasitic capacitances between
the IGBTs and the aluminium heat spreader of the IPM and the heat sink, which increases very
much the value of the generated EMI. This high EMI disturbs the proper functioning of the
electronic. By connecting the heat sink to ground (i.e. negative DC-link), a small-impedance path
is created for the parasitic currents and the generated EMI is kept therefore at low values.
5.2.3.1. Bootstrap circuit
One voltage source (typical recommended value of 15 V) will supply the LVIC circuitry. By using
bootstrap capacitors, the same voltage source is used to supply also the gate drivers of the HVIC
units, as shown in Fig. 5.9. For each inverter phase the bootstrap circuit consists of an ultra fast
recovery diode BD , a charge current limiting resistor BR and a floating supply capacitor BC . The
diode must block voltages equal to the maximal rated collector-emitter voltage of the IPM (1200V)
and must have a recovery time below 100 ns. In order to protect the circuit from dangerous
transients of the voltage supply, one Zenner-diode will be used to clamp the voltage at a certain
level. Undesired high-frequency oscillations are eliminated by low-impedance ceramic capacitors.
Before the PWM-operation starts, the low-side IGBTs are turned on, in order to allow the charging
HVIC
LVIC
=
P
U
NU
VUFB
DB
Standard charging loopfor U-phase
IGBT1
IGBT2
FDi1
FDi2
DBUltra fast recovery diode
CBFloating supply capacitor
RBCharge current limiter
ZBZenner diode
VDLVIC fixed voltage supply
VDBHVIC floating voltage supply
Ext
ern
al b
oo
tstr
ap
cir
cuit
DC-RSHUNT
iU
VUFS VDB
VN1
VNC
VD
RB
ZB
CB
Figure 5.9 The bootstrap circuit for one IPM-leg
Industrial fieldbus solution for motion coordination and monitoring 59
of the bootstrap capacitors. In Fig. 5.9, during the PWM operation, the voltage of the capacitor BC
will float depending on the potential of the terminal UFSV . When IGBT1 is turned on, the bootstrap
voltage DBV will decrease due to the current consumed by the drive circuit. Before turning on
IGBT2, there will be a dead time when both switches are turned off. If the motor current in the U-
phase is positive, than the diode FDi2 will conduct and the bootstrap capacitor can recharge. For a
negative phase current, DBV will continue to decrease as the diode FDi1 is conducting. When IGBT2
is turned on the bootstrap voltage will increase again.
5.2.3.2. Current measurement and over-current protection
The current in each motor phase was measured by shunt resistors, which are connected between
the emitters of the low-side switches and the negative rail of the DC-link. Their advantages in
comparison with hall-effect sensors are the low price and the linearity within a wide range of
frequency spectrum. The following aspects must be considered though when using this
measurement method:
The current can be measured only if the low-side switches are turned on. For the common
PWM techniques this is usually done during the zero voltage vector 0U (Fig. 2.5)
The modulation method is important for the accuracy of the measurement at high
switching frequencies and high modulation indexes
It is not possible to detect eventual short circuits between the motor phases and the
grounded motor enclosure
The ground of the measurement circuitry is connected to the negative rail of the DC-link
The shunts must have small tolerance, very low inductance and should be stable with
temperature variations
The star connection of the motor phases allows the measurement of only two phase-currents
because the third is redundant. The measurement is possible only when the low-side switches are
turned on and at modulation indexes near to one, an accurate measurement can fail due to the
transients of the measurement circuit. Using a shunt for each phase allows the measurement of
those two phase-currents for which the low-side switches have the largest turn-on times. Even so,
there are some values of the voltage space vector for which a precise current measurement is still
difficult, as shown in Fig. 5.11.
Regarding the modulation methods, these can be split in two main categories:
Continuous modulation methods, where a switching action occurs in each inverter leg at
every sampling interval. These methods are using both zero voltage vectors 0U and 7U
Discontinuous modulation methods, where switching actions occur only in two inverter
legs at every sampling interval. These methods are using only one of the zero voltage
vectors, so that the non-switching inverter leg is clamped either to the positive or the
negative DC-link
The assessment of the modulation methods is made in the literature [27], [55] by means of the
produced current distortion for different modulation indexes. For modulation indexes smaller
than 0.8, the total harmonic power losses are smaller for continuous modulation methods, as
Industrial fieldbus solution for motion coordination and monitoring 60
those presented in the Fig. 2.7, compared with discontinuous modulation methods (e.g. 60° and
120° Flat-Top). For higher modulation indexes the situation reverses. Regarding the harmonic
power losses, the average value of the modulation index is therefore a factor in deciding upon the
most appropriate modulation method. In applications where the value of the common-mode
voltage is of big concern, discontinuous modulation methods are more appropriate. For this
experimental set-up the modulation shown in Fig. 2.7b was used (continuous modulation method).
The conventional space vector modulation is a continuous modulation method, where the phase
of the reference voltages Sarefu , Sbrefu define the active sector and their amplitudes decide upon
the turn-on period of the switches. For the carrier based continuous modulation methods the
commutation times for each inverter leg are calculated easy from the reference phase voltages.
Two such methods were presented in the second chapter in Fig. 2.7. For these methods, by adding
the offset voltage component ofsu to the inverter’s output voltages, the potential of the star point
0Su is modified without affecting the phase voltages (defined in the Figs. 2.5 and 2.6). The upper
index md in the equation (5.1) denotes the modified value of a variable due to ofsu .
The mean value of the inverter’s output voltage over a sampling period will be determined by the
turn-on period onYt of the low-side switches:
U2 U7
UD/2
U0 U1 U2 U1 U0
S1
S2
S3
t
t
t
T0
-UD/2
UD/2
-UD/2
UD/2
-UD/2
UD/2
-UD/2
UD/2
-UD/2
UD/2
-UD/2
S1
S2
S3
t
t
t
T0
UD/2
-UD/2
UD/2
-UD/2
UD/2
-UD/2
U1 U0U0 U1 U2
a) Continuous modulation b) Discontinuous modulation
Figure 5.10 Modulation methods
1 10 20 30 1
10 20 30
10 20 300 10 20 30 0
2 1 1
3 3 3
1
3 3
mdS ofs ofs ofs S
md md md
mdS ofs ofs ofs ofs S ofs
u u u u u u u u
u u u
u u uu u u u u u u u u u
(5.1)
0 0
0 0
00 0
(Y 1, 2, 3)1 1
; 2 2 2
1 ;
2
onYD DY onY onY D
YonY Y SYref ofs
D
tU Uu t T t U
T T
ut T u u u
U
(5.2)
Industrial fieldbus solution for motion coordination and monitoring 61
The maximal turn-on periods onYt of the low-side switches at maximal modulation index is:
Carrier based modulation (Fig. 2.7a and equation (2.22))
Carrier based (Schörner [28]) modulation (Fig. 2.7b and equation (2.23))
The equations 5.3 and 5.4 show that for the current measurement at unity modulation indexes the
Schörner modulation is more adequate than the modulation of Fig. 2.7a, because the maximal
turn-on periods of the low-side switches are significantly larger. At maximal modulation index,
the Schörner modulation becomes discontinuous. As shown in Fig. 5.11, the phase currents are
always measured in the middle of the zero-voltage vector 0U , at the beginning of each control
cycle. Therefore the maximum allowed oscillation time of the measurement circuit osct is half of
the maximal turn-on period of the low-side switches. The maximal value of osct is 3.35 μs for the
modulation of Fig. 2.7a and 6.7 μs for the Schörner modulation. These are ideal values though
because the dead-time effect was neglected. As shown in the appendix, the IPM requires a very
large dead-time, namely 3.3 μs, so that the required value of osct becomes 3.4 μs for the Schörner
modulation A reliable measurement isn’t possible in this case with the modulation of Fig. 2.7a.
1 1,min 2 3
2 1,min
1,max 0 0
2,max 3,max 0
; 3 2 3
2 4 3
1 4 3 3 93,3 ; 1002
1 4 3 2 3 6,72
D DS ref S S ref S ref
S ref S Dofs
D D
onD
D D
on onD
U Uu u u u
u u UU
U U
t T s T sU
U U
t t T sU
(5.3)
,max1 1,min 2 3
,max
1,min
1,max 0 0
2,max 3,max 0
3 ; ;
23 2 3
23 3
1 2 3 3 100 ; 1002
1 2 3 2 32
D DSS ref S S ref S ref D
S D Dofs S
D D D
onD
D D D
on onD
U Uu u u u u U
u U UU u
U U U
t T s T sU
U U U
t t TU
0
31 13,4
2T s
(5.4)
Industrial fieldbus solution for motion coordination and monitoring 62
The measured value for osct is 4.9 μs (Fig. 5.12). This value exceeds with 1.5 μs the required value.
An easy solution to avoid this inconvenient is to start the analog-to-digital conversion not at the
beginning of the interrupt service routine, but with a fix delay of minimum 1.5 μs.
Conclusion: As known from literature [27], [55], discontinuous modulation methods are
producing, for high modulation indexes (> 0.8), less harmonics than the continuous modulation
methods (assuming a 50% increase in the switching frequency for the discontinuous modulation
methods). For lower modulation indexes, the continuous modulation methods are offering better
results. In this set-up, the Schörner modulation was used, as it offers the largest turn-on periods
of the low side switches among the continuous modulation methods.
b
a
[1 0 0][0 1 1]
[0 0 1] [1 0 1]
[0 1 0] [1 1 0][S1 S2 S3]
uSmax
uSmax
[1 1 1]
[0 0 0]
S1
S2
S3
t
t
t
S1
S2
S3
t
t
t
T0
ton1,max = 6,7µs
Currentsampling
points
tosc = 3,35µs
Currentsampling
points
Modulation of Fig. 2.7a
Schörner modulation (Fig. 2.7b)
uSmax
T0 = 100µs
ton1,max = 13,4µs
tosc = 6,7µs
T0 T0 = 100µs
UD/2
-UD/2
UD/2
-UD/2
-UD/2
UD/2
-UD/2
UD/2
-UD/2
-UD/2
UD2
3 *)
See remark on page 14*)
U1
U2U3
U4
U5 U6
U0
U7
Figure 5.11 Characteristics of the modulation methods at unity modulation index
IGBT2 Control Signal(low-side IGBT of phase U)
Voltage input at the ADC(positive phase current)
tosc
ton1,max
dead
-tim
e
Figure 5.12 Characteristics of the current measurement circuit at unity modulation indexes
Industrial fieldbus solution for motion coordination and monitoring 63
The voltage drop on the shunt resistances will be amplified, level shifted and then digitally
converted, as shown in Fig. 5.13. The shunt resistance is rated at 35 mΩ/3 W, has a tolerance of
0.5% and an inductance smaller than 3 nH. A positive current value of +12.5 A will correspond to
an input voltage of 3 V at the ADC and a negative value of -12.5 A will correspond to 0 V. This is
the maximum conversion range of the 12bit ADC.
The main controller of the DCU, the TMS320F2812 DSP from Texas Instruments, will turn off all
switches if one of the phase currents reaches ±12.5 A. Beside this software protection, a hardware
protection based on comparators was implemented. Two comparators are used for each inverter
leg to detect both positive and negative currents. An overcurrent is detected around the value of
±15 A. The outputs of all six comparators are forming an AND-connection. This is connected to the
INC terminal of the IPM over a NAND-gate (Fig. 5.13). The input of the comparators has to be low-
pass filtered in order to avoid undesired fault conditions due to noise. The time constant of this
filter was set at 1.5 μs. Fig. 5.14 shows the reaction of the protection circuit in case of an
overcurrent. The active-low fault signal is connected with the PDPINT terminal of the mentioned
DSP, which assures that all switches are turned off.
NW
RSHUNT-+
DC-(GND)
ADCChannel
R1
R1
R2
R2
C2
C2
VREF
1.5V
3.3V
VNC
CINPS22056
-
-
+
+-0.5V
+0.5V
RF
CF
5V
5V
5V
RP
DifferentialAmplifier
Analog-to-DigitalConverter
Comparators
NANDFilterUIW
0...3V
Figure 5.13 Current measurement and over-current protection for one inverter leg
UR
IGBT1 (control signal)
IRFO
(active low)
P
U
IGBT1(ON)
IGBT2(OFF)
DC-
IGBT3
IGBT4(ON)
RSHUNT
V
IRUR
PWMUnit
(OFF)
PDPINTLogic
6
ControlSignals
FaultSignal
FO
VSIDCU
eZD
SP
F28
12eZ
DS
PF
2812
Figure 5.14 Overcurrent protection test for one inverter leg
Industrial fieldbus solution for motion coordination and monitoring 64
5.2.3.3. Dimensioning of the DC-link capacitors
The DC-link of a VSI contains four aluminium-electrolytic capacitors, connected as shown in Fig.
5.15, the corresponding voltage-sharing resistors and a snubber capacitor. The electrolytic
capacitors are used to compensate the difference between the power requirement of the inverter
and the output power of the diode bridge rectifier. They supply the input current of the inverter
with pulse frequency and are a source of transient power peaks. The instantaneous value of the
current flowing through the DC-link capacitors is the difference between the AC-component of
the rectifier’s output current and AC-component of the inverter’s input current.
In [56] was shown that the current stress of the DC-link capacitors can be analytically determined
in the time domain with sufficient accuracy for sinusoidal inverter output currents and constant
DC-link voltage. Due to the symmetry of the ideal three-phase system and due to the phase-
symmetric structure of the inverter, the current analysis was limited to a 60° interval of the
inverter’s output voltage fundamental period. By this, the RMS-value of the DC-link capacitor
current is:
where 2 Nf t and Nf is the output frequency of the inverter. Inserting (5.5) in (5.6) yields to:
As the terms ,D aci and aci do not contain harmonics in the same frequency range, the integral term
of (5.7) can be neglected, so that:
The current contribution determined by the mains-commutated input rectifier depends on the
imposed voltage ripple of the DC-link voltage. The output voltage of the bridge rectifier has a
period of 300 Gf Hz . Inside this period the DC-link capacitors will be loaded during the time:
The difference between the maximal DC-link voltage ,maxDU and the minimal value ,minDU defines
the voltage ripple. The current peak value during the charge time is:
,C D ac aci i i (5.5)
2 /32 2, /3
3 C RMS CI i d
(5.6)
2 /32 2 2, , , , ,/3
3 C RMS D ac RMS D ac ac ac RMSI I i i d I
(5.7)
2 2 2, , , ,C RMS D ac RMS ac RMSI I I (5.8)
,min
,max
50
arccos
2
D
D
Cmains
Hz
U
Ut
f
(5.9)
,max ,min
C
D Dt EL
C
U UI C
t
(5.10)
Industrial fieldbus solution for motion coordination and monitoring 65
where ELC represents the DC-link capacitance of one VSI.
The corresponding RMS-value is:
Similarly for the discharging time DCt , the corresponding RMS-value is:
The total current RMS-value from the rectifier side will be:
The worst-case current stress estimation, presented in [56] as a basis for the dimensioning of the
DC-link capacitor is:
The selected electrolytic capacitor is rated at 330 μF/400 V and has a tolerance of ±20%. The total
capacitance of the DC-link is approximately 1.5 mF. The value of the desired voltage ripple was set
at ±2 V. With these values, by using the equations (5.8)-(5.14) the current RMS-value for the DC-
link capacitors of one VSI is , 5.85 C RMSI A . The maximal continuous current supported by the
selected capacitor is 3.84 A, so that the parallel connection from Fig. 5.15 had to be used, where
each capacitor conducts the half of the current stress value ,C RMSI .
The voltage rating of the electrolytic capacitor is selected based on the DC-link voltage and
tolerance values of the capacitors. The minimal voltage value, which has to be supported by one
electrolytic capacitor, is:
Because stray inductances are present between the electrolytic capacitors and the P, NU, NV, NW
terminals of the IPM, a snubber capacitor must be used to protect the IPM against high surge
voltages. It provides a low inductance path during the switching operation and therefore is placed
as closed as possible to the IPM terminals.
5.2.3.4. The interface with the Distributed Control Unit and the electronic supply
The interface between the VSI and the DSP-board of the DCU is shown in Fig. 5.15. The current
measurement circuit is connected with the analog circuit of the DSP-board. On the digital side of
the interface, voltage transceivers are used. They act like line drivers and translate also the
voltage between the DSP-board and the IPM from 3.3 V to 5 V respectively and vice versa, as the
DSP is only 3.3 V compatible. The electronic supply of a servo-controller unit is realised by an
AC/DC converter, which has to isolate, in continuous operation mode, the voltage / 2DU . Voltage
regulators are used to obtain other necessary voltage levels than those generated by the AC/DC
2,C Ct RMS t C GI I t f (5.11)
2,DC DCt RMS t DC GI I t f (5.12)
2 2, , , ,C DCD ac RMS t RMS t RMSI I I (5.13)
,
,2
N RMSac RMS
II
(5.14)
max
,minmax min
560 1.2336
1.2 0.8D
Cap
U tolerance VU V
tolerance tolerance
(5.15)
Industrial fieldbus solution for motion coordination and monitoring 66
converter. In case of a failure of the AC/DC converter or of an interruption of the mains’ voltage
during normal operation, the bootstrap circuit as well as the DC-link capacitance are still charged.
The bootstrap voltage remains above the under voltage threshold value of the IPM for few
milliseconds during which short circuits could occur. A single voltage transceiver is used for the
six control signals (+3.3 V to +5 V translation). In order to avoid possible short circuits in this case,
pull-down resistances (not shown in Fig. 5.15) are used for the outputs of this transceiver, which
avoid erroneous switching if the control signals from the DSP are floating or if they are in high
impedance state.
5.3. The Information Processing Unit
At the IPU-level there are four DCUs and the CCU. One DCU together with a VSI builds a modular
servo-controller.
5.3.1. The Distributed Control Unit
The DCU contains a main controller and the associated communication interfaces:
The position sensor interface
The interface with the adjacent DCUs (SPI-interface)
The interface with the CCU (EtherCAT-interface)
+
-+
+1.5VREF
+3.3V
OUT
OUT
-+
+0.5V
+5V
-+
+15V
-15V
-0.5V
+5V
Pull-Up
INV.
+
Pull-Up
+5V
+15V +3.3V
+5V
+5V +3.3V
+5V
+3.3V
+5V+15V-15V
ACDC
VoltageRegulator
DC +
DC -
P
U
V
W
NU
NV
NW
IPM PS22056
eZD
SP
F2
812
Co
ntr
olle
r B
oard
An
alo
g C
ircu
it (
AD
C)
Shunt U
Shunt V
Shunt W
NU
NV
NW
PD
PIN
T
Current Measurement Circuit
External Overcurrent Protection
3X
N
Bootstraping &Signal Conditioning
GND
3X
3 x AND
+
+
+
+C1
C2
C3
C4
Rvs
Sn
ub
be
r
VUFB
VUFS
VP1
UP
VVFB
VVFS
VP1
VP
VWFB
VWFS
VP1
WP
VPC
VN1
UN
VN
WN
FO
CFO
CIN
VNC
+
+
IN
IN
NU
NV
NW
L3
Electronic Supply Circuit
+15V +1.5VVoltageRegulator
iD,ac
iC/2
+15VGND
GND
Rvs +5V +3.3V
+5V +3.3V
+5V +3.3V
iac
Figure 5.15 Structure of the VSI and the DCU interface
Industrial fieldbus solution for motion coordination and monitoring 67
The main controller is the TMS320F2812 DSP from Texas Instruments, which has a fixed-point
arithmetic and a maximal processing frequency of 150 MHz. Its cost and characteristics of the CPU
and peripheral units makes it a suitable solution for servo-drive applications [57]. The DSP-board
eZDSPF2812 was selected for the set-up. Its main characteristics are shown in table 5.2.
The virtual floating-point engine IQmath from Texas Instruments was used in order to avoid the
difficulties due to the fixed-point arithmetic. The IQmath routines for e.g. computing
trigonometrical functions are executed very fast and with a very good accuracy [58].
The AD converter module has a total of 16 channels, equally separated in two sequencers, a 12-bit
converter and two dual sample and hold units. The output voltages of the differential amplifiers,
corresponding to the motor phase-currents, are connected to six AD-channels, as shown in Fig
5.16. It is possible like this to measure simultaneously any two phase-currents for which the low-
side switches have the largest turn-on times. At the end of a control cycle a pair of channels, that
is to be converted in the next cycle, is selected. The start of conversion is hardware-triggered with
a delay of 1.5 s after the beginning of a control cycle.
The high gain and offset errors of the ADC unit had to be first compensated before using the DSP-
board for the current measurement.
The DSP has two identical event manager (EVM) modules. One module includes general-purpose
timers, a PWM unit, a capture unit and a quadrature-encoder pulse (QEP) unit. The PWM unit can
generate three independent output pairs with programmable dead-time and output polarity.
Main board components Functionality
CPU
150 MHz, 32-Bit Harvard Bus-
Architecture, 32x32 Bit MAC
Operations, C/C++ Code
Compilers, Fixed-point
Arithmetic, 8 Level Pipeline
Fast code execution. The initialisation
procedures, the sensorless field oriented control
and the communication tasks are executed in
less than 100 s.
IEEE 1149.1 JTAG Interface Connection with the JTAG Controller.
External Memory Interface
(XINTF)
Connection with the EtherCAT ASIC and other
external memories.
126 kW internal Flash-memory Stores the code of the embedded system.
EVM-PWM Inverter control with integrated dead-time.
EVM-QEP Decoding of the signals from the position sensor.
AD Converter Module Simultaneous sampling of two phase-currents.
TM
S320
F281
2 Co
ntr
olle
r
Per
iph
eral
Un
its
SPI, McBSP (SPI-Compatible) Direct connection between two adjacent DCUs.
64 kW external memory Storage of different state variables.
JTAG Controller (Connection with the PC) Program debugging and internal Flash writing.
eZD
SPF2
812
Boa
rd
I/O Analog and Digital Ports Access to almost all pins of the controller.
Industrial fieldbus solution for motion coordination and monitoring 68
5.3.1.1. The interface with the position sensor
The incremental position sensor mounted on the vehicle delivers the position by means of two
symmetrical digital signals, which are phase-shifted by 90°. These signals are transmitted over the
same RS485 lines to all four DCUs, as shown in Fig. 5.17.
It is the QEP circuit of the DSP, which is responsible for decoding these signals. Because the
electronic of the servo-modules is tied to the negative rail of the DC-link and the sensor unit is
connected to protection earth (PE), the sensor interface must be electrically isolated. This was
realized with digital isolators. A PE line and a 5 V line are also transmitted along the sensor data
cable, in order to supply the local RS485 receivers and digital isolators.
5.3.1.2. The interface with the adjacent Distributed Control Units
As presented in the third chapter, the modular servo-controller must be able to communicate
with two adjacent modules. For this task the two SPI compatible units of the DSP were used:
The SPI unit
UIU
A5 A6 A7
B5 B6 B7
A0
B0
A1 A2 A3 A4
B1 B2 B3 B4
Analog MUX
AD ChannelsSequencer A
AD ChannelsSequencer B
2xSample/Hold
12-bitADC
Start ofConversion
ConversionControl
ADCClock
ConversionResult
Pair SelectAX
BX
UIV UIW
Outputs of the differential amplifiers
Figure 5.16 Simultaneous sampling and conversion of any two phase-currents
ACDC
L N
PE
+5V*(Isolated)
SensorModule
EN
Tx
Tx
Rx Rx Rx Rx Rx Rx Rx Rx
Digital Isolator Digital Isolator Digital Isolator Digital Isolator
VoltageTransceiver
DSP (QEP)
DSP (QEP)
DSP (QEP)
DSP (QEP)
EN EN EN EN
PE
+5V*
+5V +5V +5V +5V
GND GND GND GND
A
B
DC
U 1
DC
U 2
DC
U 3
DC
U 4
VoltageTransceiver
VoltageTransceiver
VoltageTransceiver
Figure 5.17 Interface between the DCUs and the sensor unit
Industrial fieldbus solution for motion coordination and monitoring 69
The multi-channel buffered serial port (McBSP) unit.
The maximal data transfer speed allowed by the DSP-board is 12.5 Mbps, which fully satisfies the
bandwidth demands in this case. At the physical layer the RS485 differential data transmission
protocol is used, as the data transmission must be immune to the switching noise and the RS485
transceivers are cheap and robust. Each DCU contains six transmitter and four receiver ports. The
SPI data and clock signals can be transmitted or received only to or from one neighbour DCU at a
time, in order to avoid short-circuits on the communication lines. All receiver ports and two of
the transmitter ports are always enabled. The other four transmitter ports, responsible for the
data and clock signal transfer to the adjacent modules, are enabled/disabled by external signals
coming from the corresponding adjacent DCUs (Fig. 5.18).
The transmit-enable input and output signals from table 5.3 are GPIO signals and are not part of
the SPI protocol.
5.3.1.3. The interface with the Central Control Unit
The CCU communicates with the DCUs over the EtherCAT fieldbus. The standardized EtherCAT
protocol is completely hardware implemented and there are now many ASICs available, which
GND
+5V
+5VGND
+5VGND
Data
Clock
Tx-Enable signal
Clock
Data
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Tx
Rx
Rx
Rx Rx
Rx Rx
Rx
Rx
Tx-Enable signal
Data
Clock
Tx-Enable signal
Tx-Enable signal
Clock
Data
DCU „n-1“
DCU „n“
DCU „n+1“
EN
EN
EN
EN
EN
EN
EN
EN EN
ENEN
DSP
EN_IR
DAT_OR
Vol
tage
Tra
nsce
iver
DAT_OL
EN_IL
CLKI
DATAI
CLK_OL
EN_OL
EN_OR
CLK_ORV
olta
ge T
rans
ceiv
er
EN_IR
DAT_OR
DAT_OL
EN_IL
CLKI
DATAI
CLK_OL
EN_OLEN_OR
CLK_OR
Vol
tage
Tra
nsce
iver
DSP
EN_IR
DAT_OR
DAT_OL
EN_IL
CLKI
DATAI
CLK_OL
EN_OLEN_OR
CLK_OR
DSP
Vol
tage
Tra
nsce
iver
Figure 5.18 SPI-based interface for communication with the adjacent modules
DAT_OR, CLK_OR SPI data and clock output signals towards the DCU on the right side DAT_OL, CLK_OL SPI data and clock output signals towards the DCU on the left side EN_IR, EN_OR Transmit-enable input and output signals from/to the DCU on the right side EN_IL, EN_OL Transmit-enable input and output signals from/to the DCU on the left side DATAI, CLKI SPI data and clock input signals
Table 5.3 Signal description of the SPI-based interface
Industrial fieldbus solution for motion coordination and monitoring 70
implement it. For this application the slave controller ET1100 was used to transform the DCU in an
EtherCAT slave unit [59]. It has three different process data interfaces (PDI):
A 16/8-bit asynchronous microcontroller interface
An SPI interface
A 32-bit digital I/O interface
Only one PDI interface can be used at a time. The EtherCAT slave-board FB1111-140 from Beckhoff
offers an easy access to the microcontroller interface of this ASIC and was used here for a fast
integration of EtherCAT in the structure of the DCU. The DSP sees the ESC as a common memory
unit. The external memory interface of the DSP allows the connection of up to three different
memory units. One memory unit is the 64 kWord RAM already available on the DSP-board, the
second unit is the ESC and the third unit is a 128 kByte Flash chip, which serves as an additional
non-volatile memory for general data. These three memory units share the same non-multiplexed
address and data buses and are accessed by the DSP over different chip-select signals. The
EtherCAT interface along with the other two memory interfaces is shown in Fig. 5.19. The
abbreviations in Fig. 5.19 are explained in the Table 5.4.
VoltageTransceiver3.3V 5.5V
Line Driver
ESC(EtherCAT
Slave Controller)
ExternalRAM
(on board)
External
FlashD[15:0]
D[7:0]
A[16:0] A[16:0]
D[15:0]
D[7:0]
CS (Flash)
CS (ESC)Ready
LatchSync
WRRD
A[16:0]
CS (XRAM)
DSP
A[12:0]
ExternalMemoryInterface(XINTF)
EventManager
eZDSPF2812
NMI
Line Driver
Line Driver
Line Driver
Line Driver
WRRD
Figure 5.19 The EtherCAT and memory interfaces for the DSP
CS Chip-select signal (active low) WR Memory write-access signal (active low) RD Memory read –access signal (active low)
Ready Signal used by the ESC to extend the active time of the read/write access NMI Non-maskable interrupt signal used by the ESC for high-priority events (active low)
A[16:0] Address signals D[15:0] Data signals Latch Signal used by the ESC to time-stamp different DSP actions Sync Signal generated at specific system times according to an internal clock of the ESC
Table 5.4 Signal description of the EtherCAT interface for the DSP
Industrial fieldbus solution for motion coordination and monitoring 71
The external Flash is 5 V compatible. The ESC and the DSP are 3.3 V compatible. A voltage
transceiver is required for the data lines between the DSP and the external Flash. For the address
lines between them, a voltage translation wasn’t necessary because the TTL inputs of the Flash are
compatible with the LVTTL outputs of the DSP. The signal wiring between the DSP board and the
ESC board reaches lengths of up to 30 cm and is situated in the proximity of the IPM, being
therefore exposed to high EMI values. Line drivers had to be implemented to amplify the power of
these signals.
The process memory of the ESC has 8 kByte, so that 13 address lines were used to access it.
Most part of the EtherCAT interface is implemented using the external memory interface of the
DSP. The last two signals of the table 5.4, Latch and Sync, are not part of this XINTF, as shown in
Fig. 5.19. Latch is a general-purpose output signal and Sync is connected to the capture unit of the
EVM. Usually these signals are dedicated to the “Distributed Clock” unit of the ESC.
At the physical layer, the interface between the DCUs and CCU is specified by the IEEE 802.3
standards, which implies voltage isolation between the communication nodes by pulse
transformers. These transformers are typically rated for an isolation voltage of 1.5 kV. This is also
the case for the network board of the EtherCAT master and the above mentioned EtherCAT slave-
board. This voltage is of course sufficient for the connection between two adjacent modules,
which have the ground potential connected to the same negative rail of the DC-link. But because
the potential of the EtherCAT master unit (PC) is at PE, the connection to the first EtherCAT slave-
unit must guarantee the same isolation voltage level as the IPM, namely 2.5 kV. This value
depends on the voltage blocking value of the IGBTs [60] and corresponds to the standard IEC1287.
For this reason, a simple network-isolating board having a 4 kV pulse transformer was
implemented as shown in Fig. 5.20, in order to connect the master PC with the first EtherCAT
TX+
TX-
RX+
RX-
RJ45
Sys
tem
Gro
un
d o
f th
eE
the
rCA
T m
ast
er
Eth
erC
AT
ma
ste
r sh
ield
75Ω
75Ω
75Ω
75Ω
1MΩ
1MΩ
10nF
10nF
TX+
TX-
RX+
RX-
RJ45S
yste
m G
rou
nd
of
the
first
Eth
erC
AT
sla
ve
Eth
erC
AT
sla
ve s
hie
ld 75Ω
1MΩ
1MΩ
10nF
10nF
Master network-board side
75Ω
75Ω
75Ω
Slave controller-board side
75Ω
75Ω
75Ω
75Ω
75Ω
75Ω
75Ω
75Ω
RJ45 RJ45
4kV Network isolator
PHY PHY
Ground potential of themodular servo-controller
-280 V
VirtualGround
VirtualGround
PE
GND
PE GND 1.5kV1.5kV
Figure 5.20 The implemented network isolator
Industrial fieldbus solution for motion coordination and monitoring 72
slave unit in the system. The centre taps of the pulse transformers are connected to a virtual local
ground over 75 Ω terminators in order to reduce the common-mode interference on the two pairs
of active wires. The unused pairs of signals are tied directly, also over 75 Ω terminators, to the
same virtual ground. The virtual ground and the shield are usually connected to PE over a 10
nF/500 V capacitor in parallel to a 1 MΩ resistance. For the first EtherCAT slave unit, the isolating
property of the 1.5 kV on-board transformer is eliminated by a direct connection of the negative
ground potential (-280 V) with the shield. Only the 4 kV transformer will support therefore
eventual voltage stresses. This additional transformer will insert obviously more power losses on
the transmission path and stress on the PHYs. The insertion-loss value as well as the return-loss
and cross-talk values have been determined for the topology in Fig. 5.20 according to the
IEEE802.3 specifications and they are not exceeding the imposed values.
The block diagram with the structure of the modular servo-controller is shown in Fig. 5.21.
5.3.2. The Central Control Unit
An ordinary PC represents the CCU. It contains the software protocol stack required for the
EtherCAT master. It needs at the hardware level only a standard network board with direct
memory access (DMA) and the processor should have a low jitter (typically bellow 2 μs). No other
special boards are required. At the software level the CCU contains the main following tools:
TwinCAT, which implements the EtherCAT master functionality
The operator interface, which includes the algorithms for motion coordination and
monitoring (Visual Basic application)
The CCS (Code Composer Studio), which is used for programming and debugging of the
DSPs in the initial test phases. It is also used to download program code in the internal
Flash memory of the DSPs
MATLAB, which is used for off-line data analysis and representation
5.4. The Position Sensor Unit
The sensor unit is made of an incremental sensor module and a 2 m-long linear magnetic scale,
which is mounted along the guideway and has a fixed pole pitch of 1 mm. The sensor module
contains an anisotropic magneto-resistive (AMR) position sensor and a high-resolution 13bit
interpolator. The measuring concept is based on the AMR-effect found in ferromagnetic materials,
where the resistance of the material will change together with the angle between the current
flowing through the material and a crossing magnetic field. The sensor module is mounted on the
vehicle, so that the active head of the sensor is perpendicular and as close as possible to the
magnetic scale. In this case the current vector is constant and the movement of the vehicle along
the magnetic scale will produce the changing magnetic field. Two resistance bridges of the sensor,
shifted in space by 90°, generate sine and cosine signals, which are afterwards digitally converted
by the interpolation circuit and supplied at the output of the sensor module as quadrature
encoded signals. The resolution of the sensor module is configured to 200 increments/mm.
Industrial fieldbus solution for motion coordination and monitoring 73
This sensor unit is cheap, accurate and was easy to interface it with the DSP, but it is not an end
solution for topologies with passive vehicles. For such topologies the optical sensors are the most
common solution nowadays in industrial applications. They are very accurate but also very
expensive (price/covered length). A cost-effective alternative, namely the capacitive sensor, has
been recently investigated in [61], [62]. Beside the capacitive sensor, another relative cheap
solution is the magnetostrictive sensor. It is an absolute sensor and has the advantage that it can
measure the position also in curves, but it cannot deliver the actual position at every 100 μs [63].
OvercurrentProtection
Circuit
ShuntResistances
DC-link andSnubber
Capacitors
BootstrapCircuit
RS485TX & RX
(SPI)
Signal adaptationDifferential amplifiers
RS485RX
(Sensor)Electronic
SupplyCircuit
ESCET1100
PHY PHY
Microcontroller interface
DigitalIsolator
Voltage Transceivers 3.3V 5V
ExternalFlash
VoltageTransceivers
andLine Drivers
DSP CoreTMS320F2812
XRAM64KW
JTAGController
Event Manager (EVM) port SPI & McBSP portsAnalog port
External Memory Interface (XINTF) port
IPMPS22056
U V WP
DC-
DC-DC+ L NMotor
Connections
EEPROMTRAFO TRAFO
SensorConnection
+5V
+5V
+15V+3.3V +1.5V
+5V
+15V
-15V
eZ
DS
F2
81
2F
B11
11-1
40
GND(DC-)
PE
DCU VSISPIConnections
EtherCAT Connections
Figure 5.21 Structure of the modular servo-controller
75
6. Control results
As shown in Fig. 2.8 the position of the vehicle can be measured by means of a linear sensor or it
can be estimated by using an EMF-based method, this if the vehicle moves with a certain minimal
speed. Therefore, the EMF-based method alone cannot be used. A sensor is necessary for starting
and stopping the vehicle. By differentiating the sensor’s position, the speed of the vehicle is
obtained. The value of the speed has a big noise component and must be filtered before being used
as a feedback for the speed controller.
6.1. Sensor based control
6.1.1. Characteristic of the induced voltages
Before testing the control algorithm, it must be assured that the induced voltages in two adjacent
segments are in phase, as expected from the mechanical construction. Otherwise phase
compensation would have been necessary. Therefore, an off-line measurement was realized,
where the vehicle was moved with an approximately constant speed during the transition
between the first two segments. The voltage in the first motor phase, which corresponds to the
induced voltage Sae , was measured simultaneously for both segments (Fig. 6.1).
eS1,seg1
+
eS1,seg1
eS1,seg1 eS1,seg1
Figure 6.1 Off-line measurement of the induced voltages during the vehicle’s transition period over
the first two segments
Control results 76
Fig. 6.1 shows that the induced voltages in the first two segments are in phase and that there is a
voltage drop in the sum of the two induced voltages during the transition period of the vehicle.
Similar results were obtained for the other two segment transitions.
6.1.2. Position control of the vehicle
The following experimental results show the controlled position over the four stator segments.
The vehicle starts from the initial position on segment 1S and crosses over segment 2S , where it
reaches the first position reference at 700 mm. Then it crosses over to the segments 3S and 4S
respectively, where it finally reaches the last position reference at 1700 mm.
S S S SS S S S S
P0
P1
P2
P3 P7
S S S S S
PF
P6 P9
S S
P5
S
P8
P4
S S
P16
P17
S
P20
P21P10
P11
P12
P13
P14 P15 P19
D0
D1
D2
D3
D5
D6
D8
D9
D4 D7
S = Reading head of the position sensor mounted on the vehicle
Initial startposition
Finalposition
S1 S2 S3 S4
P18
Figure 6.2 Main vehicle positions during the controlled transition over the four segments
Position Associated action
P1 DCU1 sends a communication request to DCU2. The communication is acknowledged. P2 VSI2 begins the commutation process and DCU2 controls the q-current to zero. P3 DCU2 becomes a slave controller and controls the q-current according to the received
reference value from DCU1.
P4 The middle point between the stator segments. The mastership for position control will be transferred to DCU2 shortly after this point, when DCU1will control the q-current according to the received reference value from DCU2.
P5 DCU1 starts to control the q-current to zero. The vehicle covers only the second stator segment.
P6 VSI1 stops its commutation process. Only VSI2 is active. P7 DCU2 and DCU1 interrupt their point-to-point communication.
Distance Associated state
D1 The thrust force is produced by both VSI1 and VSI2. D2 Both VSI1 and VSI2 are active. D3 Point-to-point communication running between DCU1 and DCU2.
Table 6.1 Actions during the vehicle transition over two segments
Control results 77
As the vehicle crosses from segment 1S to segment 2S , it passes through the positions marked in
Fig. 6.2. Each position has an associated action as shown in Table 6.1. Similar actions occur also
during the other two segment transitions. An algorithm implemented in the DCUs defines the
transition process of the vehicle. This algorithm is represented by the flowchart of Fig. 6.3.
a) Flowchart of the transition process
b1) Transition between Sn and 1Sn seen from the point of view of DCUn
b2) Transition between Sn and 1Sn seen from the point of view of DCUn+1
b3) Transition between 1Sn and Sn seen from the point of view of DCUn+1
The flowchart begins with the initial state, which implies that the DCU is in the EtherCAT-state
Operational, the inverter is active and the communication with the adjacent DCU is running.
EtherCAT OperationalState & Active VSI &
Running communication
PX
STATE 1
STATE 2
RS = 4?
STATE 3
PY
STATE 4
Check for 5 cyclesthe value of RS
RS = 3?
STATE 5
PZ
P4
S
S
S S
PY
PB
PZP5 =
Corresponding position:
PB
Corresponding position:
PA
S
Sn Sn+1
P2
S S
PAPX
P3
=
S
P4
SS
Sn Sn+1
RS=3
RS=4
SS
P4
SS
Sn Sn+1S
PY
PBRS=3
PZ P5=
a) b1)
b2)
b3)
No
Yes
PX, PY, PZare state-change
triggering positions
RS = Received StateIt is the state of the
adjacent DCU
No
No
Yes
Yes
Yes
Yes
No
No
Figure 6.3 Description of the vehicle’s transition process between two stator segments
Control results 78
Fig. 6.3-b1) describes the transition between the segments Sn and 1Sn as seen from DCUn. As
the vehicle starts from segment Sn , DCUn is a master controller (State 3). When the triggering
position, PY , is detected, DCUn changes to State 4. The position PY does not coincide with the
middle point between the segments ( 4P ), but has a displacement of one millimeter. This helps to
avoid abnormal state changes. In the same control cycle when PY was detected, DCUn is still the
master controller. At the end of this cycle its actual state as well as the state-variables of its
position and speed controllers are sent to DCUn+1. In the next cycle DCUn releases the mastership
and becomes a slave controller. DCUn+1, which at this moment is in State 2, receives the state
number of DCUn and changes to State 3. DCUn checks though if DCUn+1 really become the master
controller. If DCUn doesn’t receive this acknowledgment from DCUn+1 during five control cycles,
will change to State 5. In this state DCUn will stop the vehicle and signal the CCU that an error
occurred. In case of a successful mastership exchange, DCUn will change first to State 2 and when
the triggering position, PZ , is detected, it will change to State 1.
Fig. 6.3-b2) presents the same transition but as seen from DCUn+1, which begins from State 1.
In Fig. 6.4 the reaction of the current controllers in all stator segments are shown, as well as the
controlled speed and position. For this measurement, the speed was limited to 2 m/s
(approximately half of the maximal speed) and the q-current was limited to about half of the
maximal current. A 5 ms first order digital filter was used to filter the speed obtained from the
position sensor. From the current waveforms one can recognize the transition periods where the
inverters of two adjacent DCUs are active, as well as the periods where only the inverter of one
segment remains active, as the vehicle completely leaves the other segment. The positions
marked in this figure correspond to the positions of Fig. 6.2.
From the position and speed profiles it can be seen that the transitions between the stator
segments is smooth and bump less, i.e. the initial values of the incoming controllers were set
correctly.
Fig. 6.5 shows in detail the three transition periods of the vehicle: from segment 1S to 2S , then
from segment 2S to 3S and finally from segment 3S to 4S . The state changes of the DCUs
(according to the description in Fig. 6.3) are shown for each transition.
State Description
1 The q-current is controlled to zero. 2 The DCU is a slave controller and controls the q-current according to the received
reference value. 3 The DCU is a master controller. It controls the position and speed of the vehicle. 4 Intermediate state required for the transfer of the mastership between the DCUs. 5 This state signals an error in the communication and transition process respectively.
Table 6.2 State description for the vehicle’s transition process
Control results 79
P4
P4
P4
P11
P11
P11 P18
P18
P18
P3
P5
P10
P12
P17
P19
Segment S1
Segment S2
Segment S3
Segment S4
S2 S3 S4S1
iSq1
iSq2
iSq3
iSq4
xref x
Figure 6.4 Experimental results for the position control over four stator segments
Control results 80
PY
PZPX
PY
PY PY
PZPX
PZPX
PB
DCU1DCU2 DCU2DCU3
DCU3DCU4
iSq1
iSq2
iSq2 iSq3
iSq3 iSq4
Figure 6.5 Zoom of the vehicle’s transition periods
Control results 81
6.2. Sensorless, EMF-based control
The track of the linear direct drive system contains areas where materials are processed and areas
where materials are only transported. Along the processing areas, the presence of a position
sensor is absolute necessary to achieve the desired positioning accuracy, precision and dynamics.
Outside these processing areas a sensorless, EMF-based, control is used instead in order to save
costs. This idea is illustrated in Fig. 6.6 on the basis of a simple processing line example.
Further in this subchapter the term sensorless control refers only to the EMF-based control.
The sensorless control algorithm contains, as shown in Fig. 2.9, 2.10 and 2.11, two functional
blocks. The first block implements an EMF-observer, while the second block implements the
position and speed observer. During the vehicle’s transition period, the EMFs are observed
independently by both DCUs. The master-controller receives over the SPI-bus the EMF-value
observed by the slave-controller, computes the sum of the EMFs and estimates the position and
the speed of the vehicle with the help of a mechanical model of the motor.
Starting from the equation (2.7), the estimated values of the induced voltages in the stator
reference frame ab can be simply expressed by:
DistributedServo-Controllers
StatorSegment
Vehicle
Processing Station
Area with sensorless control
PositionSensor Unit
Figure 6.6 A simple processing line example with sensor-based and sensorless control
,
,
ˆ
ˆ
SaSa Sa ref S Sa S
SbSb Sb ref S Sb S
di te t u t R i t L
dtdi t
e t u t R i t Ldt
(6.1)
Control results 82
The space vector of the real EMF-voltage is given by:
The equation (6.4) shows the dependency of the estimated induced voltages in dq -frame with
the orientation error and the position-dependent value of the flux linkage PM . The
characteristic of the flux linkage, which implies a reduction in the sum of the EMF-values of two
stator segments during a transition period, doesn’t make possible the speed estimation based on
the EMF-magnitude. This approach is popular for rotating PMSM, where for small values of (<
30°) the estimated speed can be obtained directly from the estimated value ˆsqe and the term ˆSde is
controlled to zero, in order to correct the value of . However, this is not possible for linear
segmented motors and the speed-estimation must be based on the EMF-phase.
The design of the EMF-observer starts from equation (6.1). It is a disturbance observer based on
the electrical model of the motor [64]. As it is difficult to measure the phase voltages of the motor,
the command voltages , ,,sa ref sb refu u are used instead. The accuracy of the EMF-estimation is related
therefore with the accuracy of these command voltages. Due to the inverter’s characteristics, e.g.
dead-time, on-state voltages and turn-on/turn-off times, the command voltages will never
entirely correspond to the actual values.
The dead-time effect is a very important issue for this application, as it will be explained further.
6.2.1. Dead-time effect and compensation
The dead-time is absolutely required in order to avoid a short-circuit in an inverter leg and its
value depends on the switching devices used. It is preferable to keep it as low as possible, because
the compensation of its effects is quite difficult.
The effect upon the reference output voltages of the inverter, due to the dead-time, the turn-
on/turn-off times and the on-state voltage of the switching devices, is shown in Fig. 6.8. During
the dead-time, the potential at the motor terminals is determined by the polarity of the phase
S Sa Sb
jS Sd Sq
e t e t je t
e t e e t je t
(6.2)
0
3( ) ( )
2
Sd
Sq PM
e t
e t x v t
(6.3)
a
b
d
q
eSbeSq ∆β
β
∆β
d
eSq^
q
β
eSd^
eSa
Figure 6.7 Vector diagram of the induced voltages
3ˆ ˆ ˆ( ) ( ) sin( )
23
ˆ ˆ ˆ( ) ( ) cos( )2
ˆ
sd PM
sq PM
e t x v t
e t x v t
(6.4)
Control results 83
currents. A positive current causes a voltage loss while a negative current causes a voltage gain at
the motor terminals.
where , ,, ,DEAD C ON C OFFt t t are the dead-time and the turn-on and turn-off times of the IGBT. ,di txU U
are the on-state voltages of the diode and the IGBT. The switching time 1ont is calculated according
to equation (2.20).
The average value of the voltage deviation in one control cycle, according to the sign of the phase
current, is expressed as:
The corresponding time deviation 1DVT is:
For
1 0Si For
1 0Si
10 1 , ,0
0 1 , ,
10 1 , ,0
0 1 , ,
1( )( )
2
( )( )2
1( )( )
2
( )( )2
Don DEAD C ON C OFF tx
Don DEAD C ON C OFF di
Don DEAD C OFF C ON di
Don DEAD C OFF C ON tx
Uu t t t t U
T
UT t t t t U
Uu t t t t U
T
UT t t t t U
(6.5)
UD/2
u10 ref
T1
T2
tDEAD
tC,ON
5 V
0 V
0 V
5 V
T0/2U
IGBT1
IGBT2
FDi1
FDi2
DC-
DC+
T1
T2
UD
(iS1 > 0) (iS1 < 0)
utx
u10
udi
u10
u10(iS1 < 0)
(iS1 > 0)
utx
utx
udi
udi
tDEAD
tC,ONtC,OFF
tC,OFF
T0/2
-UD/2
UD/2
UD/2
-UD/2
-UD/2
Figure 6.8 Distortion of the inverter’s output reference voltages
1
1 10
sgn( ) DVDV S D
TU i U
T
(6.6)
1 , , 0
dtDV DEAD C ON C OFF
D
UT t t t T
U
(6.7)
Control results 84
where dtU is the average on-state voltage in a control cycle of both diodes and IGBTs. Similarly,
the deviation voltages for the other two phases can be determined. According to the data-sheet of
the PS22056 IPM, a dead-time period of at least 3.3 μs is required. The dead-time generator of the
DSP was set at 3.4 μs. This large dead-time value corroborated with the high DC-link voltage and
switching frequency values, leads to very high deviation voltages with values between ±20 V.
There are many proposals in the literature about how to calculate the compensation values for the
inverter’s output voltage. Many considerations start from the idea of adding compensation
voltages to the command voltages of the current controllers [65]-[70]. Another approach was
presented in [71], where the compensation is realised by the adjustment of the symmetric PWM
pulses in every control cycle.
Remark: In spite of the large dead-time (3.4 μs) the current control loop is able to generate close
to sinusoidal current waveforms even without any compensation in the control loop. This is
shown in the measurement of Fig. 6.10a. But the large dead-time is a big problem for the EMF-
observer and a compensation is for the observer mandatory. Fig. 6.9 shows that the compensation
is used for the EMF-observer, but the current control loop needs no compensation.
The compensation methods are confronted with one or many of the following impediments:
An accurate measurement of the current polarity is difficult due to the switching noise and
the current clamping effect
The characteristics of the inverter and the motor parameters are changing along with the
operating conditions
Variation of the DC-link voltage
The method presented in [70], depends only on the electrical model of the PMSM, as it
implements a disturbance observer in order to calculate the deviation voltages. Anyhow, this
method cannot be used in this case because the observer requires the EMF values.
Here an approach similar with the ones in [67] and [69] was adopted, where the compensation
voltage is adjusted according to the load current. This is necessary because of the variation of the
UD T0 tDEAD tC,ON tC,OFF UDV
560 V 100 μs 3.4 μs 0.9 μs 0.9 μs ≈ (±20 V)
Table 6.3 Characteristics of the VSI
3
2
PWM
EMF-Observer
u
i
UCOM
uSa ref
2
3
+
uSb ref
uS1 ref
uS2 refuS3 ref uSaC
uSbCuS1C
uS2CuS3C
iS1, iS2, iS3
-UCOM
Figure 6.9 Dead-time compensation for the implemented EMF-Observer
Control results 85
turn-off time due to the parasitic capacitance of the switching devices [69].
For the measurement in Fig. 6.10 and 6.11 the vehicle was removed from the stator segment (no
EMF) and the frequency of the orientation angle was set at 10 Hz. It can be seen that the
inaccuracies in the computation of the deviation voltages are still present and do not depend on
the load-current. Both Fig. 6.10 and Fig. 6.11 show experimental results. An exact compensation
would require a precise current measurement, a precise model of the diodes and switches and DC-
link voltage measurement. The worst-case will always occur at small load-currents and small
values of the induced voltage (small speed).
iS1 iS2 iS3 iS1 iS2 iS3
uSa ref uSb ref uSaC uSbC
a) Uncompensated command voltages b) Compensated command voltages
Figure 6.10 Uncompensated and compensated command voltages for a phase current of 1 A
uSa ref
iS1 iS2 iS3 iS1 iS2 iS3
uSb ref uSaC uSbC
a) Uncompensated command voltages b) Compensated command voltages
Figure 6.11 Uncompensated and compensated command voltages for a phase current of 2 A
Control results 86
6.2.2. The EMF observer
The EMF-observer is a disturbance observer based on the electrical model of the motor. Starting
from the equation (2.7) the equivalent linear second order system with disturbance can be defined
in matrix form as:
The disturbance model is defined by the following equations:
The term PM doesn’t depend on time and assuming that the speed variation over a control cycle
is very small, the time derivative of the induced voltages can be written as:
From (6.8) and (6.11) the model of the observer is built:
where 2I is the unity matrix. H and G are the gain matrixes of the observer and Cu is the matrix
of the compensated voltages.
The matrix Ad in the equation (6.12) implies that the estimated speed v equals the real speed v ,
which depends on the mechanical observer. The coupling between the observers is eliminated by
the assumption in the equation (6.13).
1 10 0 0
1 10 00
x u dB KA
S
S S SSa Sa Sa Sa
Sb Sb Sb SbS
S SS
R
L L Li i u edi i u eRdt
L LL
(6.8)
3( ) ( ) ( ) sin( )
23
( ) ( ) ( ) cos( )2
Sa PM
Sb PM
e t x v t
e t x v t
(6.9)
3( ) ( ) cos( ) ( )
2
3( ) ( ) sin( ) ( )
2
Sa PM Sb
Sb PM Sa
e t v x v v e t
e t v x v v e t
(6.10)
0
0d
Ad
Sa Sa
Sb Sb
ve ede edt
v
(6.11)
ˆ ˆ ˆ0
ˆ ˆ ˆ0 0 C 2
x x x xA K B Hu I
Ad Gd d d dd
dt
(6.12)
11 11
22 22
0 01 0
0 00 12 CI ; H ; G ; u SaC
SbC
uh g
uh g
Control results 87
The equations describing the observer become then:
The correction term of the observer is based on the difference between the measured and the
estimated current. In order to simplify the model, the observer’s gains are selected as: 11 22h h
and 11 22g g . Now the values of these coefficients are determined by matching the coefficients of
the characteristic polynomial of (6.12) with the negative roots of a 4th order polynomial. The roots
are chosen as real, double roots and the gains of the observer are selected as in [72]:
where 1 2,p p are the roots of the mentioned polynomial.
The freedom in selecting the values of 1 2,p p is limited by the value of the orientation error
produced by the assumption in (6.13). The orientation error should be 30 , otherwise a
significant decrease in the amplitude of the electrical force occurs. As deduced in [72], the
limitation of the orientation error is set by:
Equation (6.16) gives the ratio between the gain factors 11h and 11g , according to an imposed limit
of the orientation error (here set at 25°). The gain 11h defines the dynamic of the observer.
Because the quality of the command voltages ,SaC SbCu u is strongly affected by noise, the dynamic
of the observer had to be diminished in order to increase the filtering effect of the observer.
After the estimation of the induced voltages, the estimated orientation angle was computed as:
The term ˆSae is equal to ˆSae when the vehicle is completely inside a stator segment and with the
sum of both induced voltages, i.e. , , 1ˆ ˆSa n Sa ne e during the vehicle’s transition between the
segments Sn and 1Sn . The same applies for ˆSbe .
The sensorless method was tested on two segments, as shown in Fig. 6.12.
ˆ 0
ˆ 0Sa
Sb
ededt
(6.13)
11
22
11
22
ˆ ( ) 1 1ˆ ˆˆ( ) ( ) ( ) ( ) ( )
ˆ ( ) 1 1ˆ ˆˆ( ) ( ) ( ) ( ) ( )
ˆ ( ) ˆ( ) ( )
ˆ ( ) ˆ( ) ( )
Sa SSa Sa SaC Sa Sa
S S S
Sb SSb Sb SbC Sb Sb
S S S
SaSa Sa
SbSb Sb
di t Ri t e t u t h i t i t
dt L L L
di t Ri t e t u t h i t i t
dt L L L
de tg i t i t
dtde t
g i t i tdt
(6.14)
11 22 1 2
11 22 1 2
( )h h p p
g g p p
(6.15)
1 11
11
tanh
vg
(6.16)
ˆˆˆ
Sa
Sb
e
e
atan
(6.17)
Control results 88
The vehicle starts from the initial potion 0P and is accelerated until the position PS is reached,
when the sensorless control is activated. At this position the speed of the vehicle is big enough to
assure a reliable EMF-estimation. The sensorless motion of the vehicle ends at the position PR .
Between the positions PC and PD , the EMF-values in both segments are added to build the
values of ˆ ˆ,Sa Sbe e . The positions 3P and 5P are defined in the table 6.1. The flowchart of the
transition process from Fig. 6.3 remains valid for the sensorless control too.
The Fig. 6.13 shows the estimated EMF-values and orientation angles for both stator segments. For
the speed control the position sensor was used. In the interval defined by the positions PC and
PD , the magnitude reduction of ˆ ˆ,Sa Sbe e during the transition period can be seen. In the control
cycle when the mastership is changed between the two DCUs, the initial value of the angle 2ˆ
E is
set to the actual value of the angle 1E . The estimated angles 1 2ˆ ˆ,E E , and therefore the
estimated position ˆEx , have a high noise level, especially during the transition period.
6.2.3. The speed and position observer
The speed and position observer (mechanical observer) is designed starting from the mechanical
model of the motor, which is defined as:
where Cb is the friction coefficient and LF .is the load force.
The block diagram of the observer is shown in Fig. 6.14. It is defined by the following equation:
S
S
S
PS PC
P5
S1 S2
P0 P3
S SS
S
PRPD
Sensorless area
AddingEMF
Initial startposition
S
Figure 6.12 Sensorless transition between two adjacent stator segments
0 0 0 0
1 10
0 1 0 0
L L
Cth
v v v
F Fbd
v v Fdt M M M
x x
(6.18)
1
2
3
0 0 0 0ˆ ˆ ˆ1 1
ˆ ˆ ˆ0 0 0 1
ˆ ˆ ˆ ˆ0 1 0 0
L L L L
Cth
v v vM M E M
F F k F Fbd
v v F k v vdt M M M
x x k x x
(6.19)
Control results 89
This type of observer is usually used for speed estimation, when the position measurement has a
low resolution or a high noisy level. Here, the observer’s input position is not coming from a
sensor, but from the EMF observer and has the value ˆEx . The deviation between ˆEx and the actual
position is similar to a noisy position measurement, which has to be filtered by the observer. The
correction term of the observer is based on the difference between the position ˆEx and the
filtered position ˆMx . The gain values of the observer are determined by means of the estimation
PD
PC
Speed from the position sensor
eSa1^ +
+
+
+
ββ1E^
β2E^
PR
eSa1^ eSa2
^
eSb1^ eSb2
^
eSb1^
eSa2^ eSa1
^
eSb2^ eSb1
^
eSa2^
eSb2^
Figure 6.13 EMF-based estimation of the orientation angle (experimental results)
1
2
3
0 0
1
0 1
Av
L L
C
v v
kF F
bdv k v
dt M Mx x
k
(6.20)
Control results 90
error model of (6.20), where ˆL L LF F F , ˆv v v and ˆ ˆE Mx x x .
The characteristic polynomial of the error model is:
After matching the coefficients of the polynomial of (6.21) with a third order polynomial with the
roots 1 2 3, ,p p p , the values of the observer gain are:
The mass of the vehicle is 6.5 Kg and the friction coefficient Cb has the value of 8 Kg/s. The values
of 1 2 3, ,p p p are calculated so that they match the poles of a third order, low-pass Butterworth
filter. Due to the high noise level of the calculated position ˆEx , the cutoff frequency of the
mechanical observer must be set very low. The filtering time constant of the Butterworth filter
was set here at 15 ms. For this reason the sensitivity to the thrust force errors is high.
The experimental results shown in Fig. 6.15 show the values of the estimated speeds 1 2ˆ ˆ,v v ,
estimated orientation angles 1 2ˆ ˆ,M M and the error ˆMx x x in the position estimation, for
the sensorless control of the vehicle over two segments.
The orientation angles 1 2ˆ ˆ,M M are defined as:
-1/Mv
K1
bC/Mv
-
-+
+
++
+
+Fth
v
FL^
xE^
K2
K3xM^
Figure 6.14 The block diagram of the position and speed observer
3 2 1
3 3 2det( ) 03I Av C C
v v v
b b ks s s k s k k
M M M
(6.21)
1 1 2 3
2
2 1 2 2 3 1 3 1 2 3
3 1 2 3
( ) ( )
( )
v
C C
v v
C
v
k p p p M
b bk p p p p p p p p p
M M
bk p p p
M
(6.22)
1 1
2 2
ˆ ˆ
ˆ ˆ
M M
M M
x
x
(6.23)
Control results 91
When the vehicle is in the middle of the stator segments the following transitions occur:
The error for the position estimation is smaller than 5 mm, which means that for a pole pitch
36 mm , the error for the orientation angle ˆ 25M . When the mastership is
PS
PR
v(position
from sensor)
v1^
v2^
β2M^
β
∆x = xM - x^
β1M^
The vehicle leaves segment 2
and heads towardssegment 3
Figure 6.15 Speed sensorless control over two stator segments (constant speed)
2 1
2 1
2 1
ˆ ˆ ˆ
ˆ ˆ ˆ
ˆ ˆ ˆ
M M M
M M Mx x x
v v v
(6.24)
∆v = v - v^
β
β1M^
Figure 6.16 Orientation angle displacement and estimated speed error (zoom of Fig. 6.15)
Control results 92
changed between the two DCUs, the initial value of the speed 2v is set to the actual value of the
speed 1v . The Fig. 6.16 shows (for the same measurement from Fig. 6.15) that, as expected, the
error for the estimated speed has the biggest value during the transition period. The
commutation between the sensor-based and the sensorless control occurs at 0.6 m/s when the
estimated EMF values are accurate enough for a good position and speed estimation.
In the case presented in Fig. 6.17, the vehicle is accelerated, shortly after it completely entered the
second segment, from 1.5 m/s to 2 m/s. Also in this case the error for the position estimation
remains smaller than 5 mm. Fig. 6.17 presents experimental results.
v(position
from sensor)β
∆x = xM - x^
PS
PR
The vehicle leaves segment 2
and heads towardssegment 3
x
x1M^x2M
^
v1^
v2^
β1M^
β2M^
Figure 6.17 Speed sensorless control over two stator segments (acceleration)
93
7. Conclusions
The purpose of the work was to design a linear direct drive system for material handling and
factory automation with totally distributed power and control stages. The system was designed to
assure a high degree of modularity and scalability with minimal implementation costs. The
characteristics and the performance of the designed system were tested on a small-scale
experimental set-up. Among the multiple topologies of linear motors, the single-sided, long-stator
PMLSM proved to be the most suitable solution.
For the power stage, the trend nowadays in electrical drives is to use compact power electronic
modules with integrated gate-drivers and protection circuits. At the beginning of the work, the
selected IPM was one of the few solutions available on the market in its power class. Beside the
advantages of high-integration and low costs, the selected IPM has also some disadvantages. The
biggest disadvantage is represented by the large dead-time, which influences the dynamic and
performance of the sensorless control, as shown in chapter 6. The compact structure leads to
higher capacitance values inside the module. It was demonstrated that the electronic of the servo-
controllers, tied to the negative rail of the DC-link, functions properly under noisy conditions. A
direct connection between the inverter and the controller was therefore possible. A safe current
measurement, even at high modulation indexes, could be realised by using a shunt resistor in the
low side of each inverter leg. Depending on the average value of the modulation index, continuous
or discontinuous modulation methods can be used.
At the information processing stage, the hard real-time communication demands between the
adjacent servo-controllers was solved by the proposed point-to-point connection. Here, the
common-mode voltage problems of the RS485 devices were identified and analysed. In this case
the value of this voltage was not of concern due to the low effective inductance of the bus bar
system. The bus bar system can be of course further optimised, to reach smaller inductance
values. Solutions have been also presented in the case that the common-mode voltage exceeds
though the maximal specified values.
The task of motion coordination and monitoring was realised by means of an RTE-based fieldbus.
EtherCAT was chosen here due to its advantages presented in chapter 4. In developing the
communication structure of the system, the worst case was assumed, namely that all vehicles
could be in a transition process. This led to the necessity of implementing the SPI-based
Control results 94
communication, besides the RTE-based one. A future work will be the optimisation of the motion
coordination process with the help of special algorithms, so that the number of vehicles, which
are in a transition state at a moment, is minimal. This would allow the utilisation of only one
communication protocol, namely the RTE-based one. The overall costs would be therefore
reduced and the transient voltages on the DC-link lines would be no more of concern.
In chapter 6, an EMF-based sensorless control method was presented. The results of this control
method are satisfactory for the desired requirement, namely the sensorless transportation (with
variable speed) of materials between the processing stations. Optical sensors give the best
position measurement accuracy inside the processing stations. Their disadvantage is though their
high price. Future work is therefore necessary to compare the cost-effective solutions: the
capacitive and the magnetostrictive position sensors, and improve their accuracy.
95
Appendix
A.1 Speed control loop
The transfer function of the open control loop is:
where ,RV RVV T are the gain and the time constant of the PI speed controller and FT is the time
constant of the speed filter.
The equivalent time constant of the current control loop is approximated by:
The transfer function of the closed control loop is:
The parameters of the controller are designed by using the method of symmetrical optimum:
In order to avoid a big overshoot in the step response of the controller, a PT1 reference speed
filter was used. The pole of this filter will compensate the zero of the closed control loop.
Fth ref
-vref
-Speed
ControllerCurrent
Control Loop
Motor Model
iSq ref
kf
vf
v VRV
ReferenceFilter
MVTRV
kf
TF
TEI
iSq Fth
Fd
App. 1 Diagram of the speed control loop
1 1 1( )
1RV RV
Kf RV EI V
V T sG s
k T s T s M s
(A1)
20 0 0
34 3
2EIT D T T (A2)
20 0 0
34 3
2EIT D T T (A3)
2
4
vRV
f EI F
RV EI F
MV
k T T
T T T
(A4)
Appendix 96
A.2 Position control loop
The transfer function of the open control loop is:
where PV is the gain of the proportional controller and EVT is the equivalent time constant of the
speed control loop.
The transfer function of the closed control loop is:
According to the amplitude optimum, the gain of the controller is:
A.3 Analytical calculation of the parameters of a bi-plate DC-link bus bar system
The effective inductance of the bus bar system at low frequencies is:
-
PositionController
SpeedControl Loop
v x
TEV
vrefxref
VP
App. 2 Diagram of the position control loop
( )
1P
KEV
VG s
s T s
(A5)
0
2
1( )
11EV
P P
G sT
s sV V
(A6)
20
1 1
4 2PEV EV
VD T T
(A7)
length (l)
distance between plates (d)
width (w)
thickness (th)
App. 3 Planar bi-plate DC-link bus
0
, 02 [H]8
r heff dc r
h
t lL l
t w
(A8)
Appendix 97
where r is the relative permeability of the isolating material.
The effective inductance of the bus bar system at frequencies above 1 MHz is defined by:
where is the skin depth defined as:
is the conductivity of the conducting plates e.g. 81.68 10 [ m]Copper and f is the
frequency.
The resistance value of the two conductors at low frequencies is:
The resistance value of the two conductors at high frequencies is:
A.4 Calculation of the initial start position of the vehicle
In order to move the vehicle from standstill using the FOC, it is necessary for a PMLSM to know
the relative position between the stator windings and the initial position of the PM (vehicle).
0, [H/m]r
eff acL dw
(A9)
0
[m]f
(A10)
2dc
h
lR
w t
(A11)
4ac
lR
w
(A12)
Offset Offset
App. 4 Offset of the electrical angle of the vehicle at starting point (experimental results)
Appendix 98
The offset between the reference starting position and the real position of PM can be
approximately determined with the help of the already tuned current controllers. For this, the
reference values of both (dq) current controllers are set to zero and assuming that 0 , the
vehicle is moved with a relative constant speed. Because the currents are zero, the induced
voltage in the machine will depend only on the vehicle’s speed. The current controllers will
compensate this induced voltage, so that the voltage response Sarefu will be almost equal to the
induced voltage in the a-axis. The dead-time influence is very small due to the very low current
values. The negative zero crossing points of the voltage Sarefu will point to the real position of
PM . If these points and the information from the position sensor are known simultaneously, the
offset is determined as shown in Fig. App.4.
A.5 Characteristics of the IPM PS22056
Maximal ratings Name Symbol Condition Rating Unit
Collector-emitter voltage VCES - 1200 V IGBT collector current IC Junction temperature TJ = 25°C ±25 A Junction temperature TJ - -20 to +125 °C
Isolation voltage VISO Sinusoidal 60Hz, one minute 2500 VRMS Control input voltage VIN Between UP,VP,WP-VPC ;UN,VN,WN-VNC -0.5 to VD+0.5 V
Static and switching characteristics
Name Symbol Condition Min. Typ. Max. Unit Collector emitter
saturation voltage VCES VD = VDB = 15V
VIN = 5V, IC = 25A, TJ = 25°C - 2.7 3.4 V
Diode forward voltage VF VIN = 0, IC = -25A - 2.5 3.0 V tON 0.8 1.5 2.2 tRR - 0.3 -
tC(ON) - 0.6 0.9 tOFF - 2.8 3.8
Switching times
tC(OFF)
VCC = 600V VD = 15V, VDB = 15V
IC = 25A TJ = 125°C
Inductive load - 0.6 0.9
μs
FO pulse width tFO CFO = 22nF 1.6 2.4 - ms HVIC circuit current ID VD = VDB = 15V - - 1.3 mA
Recommended operation values
Name Symbol Condition Min. Typ. Max. Unit Supply voltage VCC Between P-NU, NV, NW 350 600 800 V
Control supply voltage VD Between VP1-VPC and VN1-VNC 13.5 15 16.5 V Control supply voltage VDB Between VU,V,WFB-VU,V,WFS 13.5 15 16.5 V
Dead-time tDEAD For each input control signal 3.3 - - μs PWM frequency fS Junction temperature TJ < 125°C - - 15 KHz Output current IRMS VCC = 600V, fS = 15KHz, VD = 15V - - 9.2 A
PWON - 1.5 - - μs Minimum PWM pulse width PWOFF Collector current IC < 25A 2.1 - - μs
Appendix 99
A.5 Photos of the experimental set-up
VehicleServo-
Controller(DCU+VSI)
CCU
DC-link bus4 KV LANMagnetics
Line Connection& Rectifier
MotorCable
DSPBoard
Eth
erC
AT
Boa
rd
Fro
mS
enso
r
Point-to-Point Connection
DC LinkCapacitors
100
Bibliography
[1] Laithwaite E. R., ”Linear Electric Motors”, M&B Technical Library, London 1971; ISBN
0263515869
[2] Boldea I., Nasar A., ”Linear Electric Actuators and Generators”, Cambridge University Press,
New York 1997; ISBN 0521480175
[3] Gieras F., Piech J., ”Linear Synchronous Motors-Transportation and Automation Systems”, CRC
Press, Florida - Boca Raton 2000; ISBN 0849318597
[4] Gurol H., ”General Atomics Linear Motor Applications: Moving Towards Deployment”, Proceedings
of the IEEE, Vol. 97, No 11, November 2009, pp. 1864-1871
[5] Hellinger R., Mnich P., ”Linear Motor-Powered Transportation: History, Present Status and Future
Outlook”, Proceedings of the IEEE, Vol. 97, No 11, November 2009, pp. 1892-1900
[6] Park D. Y., Shin B. C., Han H., ”Korea’s Urban Maglev Program”, Proceedings of the IEEE, Vol.
97, No 11, November 2009, pp. 1886-1891
[7] Gustafsson J., ”Vectus-Intelligent Transport”, Proceedings of the IEEE, Vol. 97, No 11,
November 2009, pp. 1856-1863
[8] Cassat A., Kawkabani B., Perriard Y., ”Power Supply of Long Stator Linear Motors-Application to
Multi Mobile System”, IEEE annual meeting of Industry Application Society, IAS 2008
[9] Zhu Y., Lee S., Cho Y., ”Topology Structure Selection of Permanent Magnet Linear Synchronous
Motor for Ropeless Elevator System”, IEEE International Symposium on Industrial Electronics,
ISIE 2010
[10] Perreault, B., ”Optimizing Operation of Segmented Stator Linear Synchronous Motors”,
Proceedings of the IEEE, Vol. 97, No 11, November 2009, pp. 1777-1785
[11] Henning U., Hoke D., Nothhaft J., ”Development and Operation Results of TRANSRAPID
Propulsion System”, International Conference on Magnetically Levitated Systems and Linear
Drives, Maglev 2004
[12] Mutschler P., Silaghiu S., ”Linear Drives for Material Handling and Processing: A Comparison of
System Architectures”, 34th annual conference of IEEE, IECON 2008
[13] Corsi N., Coleman R., Piaget D., ”Status and New Development of Linear Drives and Subsystems”,
6th International Symposium on Linear Drives for Industrial Applications, LDIA 2007
101
[14] http://www.parker.com/portal/site/PARKER|, “Parker Linear Motor Solutions – Catalog
8092/USA”, SectionA.pdf (February 2012)
[15] Fernandes Neto T. R., Mutschler P., ”Short Primary Linear Synchronous Motor Drive with an
Ultracapacitor Regenerative Braking System for Material Handling and Processing”, 8th
International Symposium on Linear Drives for Industrial Applications, LDIA 2011
[16] Mihalachi M., Leidhold R., Mutschler P., ”Linear Drive System for Combined Transportation and
Processing of Materials”, 35th annual conference of IEEE, IECON 2009
[17] Purcarea C., Kedarisetti J., et al., ”A Motor-Friendly and Efficient Resonant DC-Link Converter”,
13th European Conference on Power Electronics and Applications, EPE 2009
[18] Ranky P. G., ”MagneMotion’s Linear Synchronous Motor (LSM) Driven Assembly Automation and
Material Handling System Designs”, Assembly Automation Journal, Vol. 27, No. 2, 2007, pp. 97-
102
[19] http://www.magnemotion.com/industrial-automation/QS_layout.cfm, (February 2012)
[20] Zesheng D., Boldea I., et al. ”Fields in Permanent Magnet Linear Synchronous Machines”, IEEE
Transactions on Magnetics, Vol. 22, No 2, March 1986, pp. 107-112
[21] Leonhard W., ”Control of Electrical Drives”, 3rd Ed., Springer-Verlag, 2001; ISBN 3-540-41820-2
[22] Denkena B., Tönshoff H. K., et al., ”Analysis and Control/Monitoring of the Direct Linear Drive in
End Milling”, International Journal of Production Research, Vol. 42, No. 24, December 2004,
pp. 5149-5166
[23] Benavides R., Mutschler P., ”Compensation of Disturbances in Segmented Long Stator Linear
Drives using Finite Element Models”, IEEE International Symposium on Industrial Electronics,
ISIE 2006
[24] Ogasawara S., Akagi H., et al., ”A Novel PWM Scheme of Voltage Source Inverters based on Space
Vector Theory”, Archiv für Elektrotechnik, Springer-Verlag, Vol. 74, 1990,
[25] Holtz J., Beyer B., ”Optimal Pulsewidth Modulation for AC Servos and Low-Cost Industrial Drive”,
IEEE Transactions on Industry Applications, Vol. 30, No. 4, August 1994, pp. 1039-1047
[26] van der Broeck H., Skudelny H. C., ”Analysis and Realization of a Pulsewidth Modulator based on
Voltage Space Vectors”, IEEE Transactions on Industry Applications, Vol. 24, No. 1, February
1988, pp. 142-150
[27] Reinold H., ”Optimierung dreiphasiger Pulsdauermodulationsverfahren”, doctoral dissertation,
Verlag der Augustinus Buchhandlung, Aachen1996, ISBN 3-86073-235-8
[28] Schröder J., ”Bezugsspannung zur Umrichtersteuerung”, ETZ-B, Vol. 27, No. 7, 1975, pp. 151-152
[29] Leidhold R., ”Position Sensorless Control of PM Synchronous Motors based on Zero-Sequence Carrier
Injection”, IEEE Transactions on Industrial Electronics, Vol. 58, No. 12, 2008, pp. 5371-5379
102
[30] Holtz J., ”Acquisition of Position Error and Magnet Polarity for Sensorless Control of PM
Synchronous Machines”, IEEE Transactions on Industry Applications, Vol. 44, No. 4, August
2008, pp. 1172-1180
[31] Boldea I., Paicu M. C., et al., ”Active Flux Concept for Motion-Sensorless Unified AC Drives”, IEEE
Transactions on Power Electronics, Vol. 23, No. 5, September 2008, pp. 2612-2618
[32] Linke M., Kennel R., et al., ”Sensorless Speed and Position Control of Synchronous Machines using
Alternating Carrier Injection”, IEEE International Conference on Electric Machines and Drives,
IEMDC 2003
[33] Zurawski R., ”The Industrial Communication Technology Handbook”, CRC Press 2005; ISBN 978-
0-8493-3077-3
[34] Decotignie J., ”The Many Faces of Industrial Ethernet [Past and Present]”, IEEE Industrial
Electronics Magazine, Vol. 3, No. 1, March 2009, pp. 8-19
[35] Rostan M., Stubbs J. E., et al., ”EtherCAT enabled Advanced Control Architecture”, IEEE/SEMI
Advanced Semiconductor Manufacturing Conference, ASMC 2010
[36] Texas Instruments, ”TMS320x281x DSP Multichannel Buffered Serial Port Reference Guide”,
Literature Number SPRU061C, Revised November 2007
[37] Silaghiu S., Mutschler P., ”Communication Topology in a Modular Servo-Drive System based on
Long Stator Permanent Magnet Synchronous Linear Motor”, 5th IET International Conference on
Power Electronics, Machines and Drives, PEMD 2010
[38] Texas Instruments, ”RS-422 and RS-485 Standards Overview and System Configurations”,
Literature Number SLLA070D, Revised May 2010
[39] Bakran M. M., Helsper M., et al., ”Multicommutation of IGBTs in Large Inverters”, 11th European
Conference on Power Electronics and Applications, EPE 2005
[40] Bock B., ”Switching IGBTs in Parallel Connection or with Enlarged Commutation Inductance”,
Doctoral dissertation, Bochum 2005, URL http://www-brs.ub.ruhr-uni-
bochum.de/netahtml/HSS/Diss/BockBurkhard/diss.pdf
[41] Skibinski G. L., Divan D. M., ”Design Methodology and Modelling of Low Inductance Planar Bus
Structures”, 5th European Conference on Power Electronics and Applications, EPE 1993, pp.
98-105
[42] Zare F., Ledwich F. G., ”Reduced Layer Planar Busbar for Voltage Source Inverters”, IEEE
Transactions on Power Electronics, Vol. 17, No. 4, July 2002, pp. 508-516
[43] Caponet M. C., Profumo F., et al., ”Low Stray Inductance Bus Bar Design and Construction for
Good EMC Performance in Power Electronic Circuits”, IEEE Transactions on Power Electronics,
Vol. 17, No. 2, March 2002, pp. 225-231
103
[44] Decotignie J. D., ”A perspective on Ethernet-TCP/IP as a Fieldbus”, IFAC 4th International
Conference on Fieldbus Systems and their Applications, FeT 2001
[45] Thomesse J. P., ”Fieldbus Technology in Industrial Automation”, Proceedings of the IEEE, Vol.
93, No 6, June 2005, pp. 1073-1101
[46] Felser M., ”Real-Time Ethernet-Industry Prospective”, Proceedings of the IEEE, Vol. 93, No 6,
June 2005, pp. 1118-1129
[47] Decotignie J. D., ”Ethernet-Based Real-Time and Industrial Communications”, Proceedings of the
IEEE, Vol. 93, No 6, June 2005, pp. 1102-1117
[48] Zurawski R., ”Embedded System Handbook-Networked Embedded Systems”, CRC Press 2009; ISBN
978-1-4398-0761-3, Part IV, Chapter 21
[49] EtherCAT Technology Group (ETG), ”EtherCAT Communication Specification”, Member
Download Area, Version 1.0 2004
[50] Beckhoff Automation, ”The Windows Control and Automation Technology” URL
(http://www.beckhoff.com/default.asp?twincat/default.htm), Link retrieved in 2011
[51] Texas Instruments, ”Running an Application from Internal Flash Memory on the TMS320F28xxx
DSP”, Literature Number SPRA958J, Revised June 2011
[52] Texas Instruments, ”TMS320C28x Assembly Language Tools-User Guide”, Literature Number
SPRU513D, Revised May 2011
[53] Majumdar G., Iwasaki M., et al., ”Compact IPMs in Transfer Mold Packages for Low-Power Motor
Drives”, Proceedings of the International Symposium on Power Semiconductor Devices and
ICs, ISPSD 2004
[54] Motto E., Donlon J., et al., ”A 1200V Transfer Molded DIP-IPM”, POWEREX Technical Library
URL (http://www.pwrx.com/Library.aspx?s=1^0|2^65|3^0|), article published in 2004
[55] Kolar W. J., Ertl H., et al., ”Influence of the Modulation Method on the Conduction and Switching
Losses of a PWM Converter System”, IEEE Transactions on Industry Applications, Vol. 27, No. 6,
December 1991, pp. 1063-1075
[56] Kolar W. J., Round S. D., ”Analytical Calculation of the RMS Current Stress on the DC-link
Capacitor of Voltage-PWM Converter Systems”, IEE Proceedings of Electric Power Applications,
Vol. 153, No. 4, July 2006, pp. 535-543
[57] Texas Instruments, ”TMS320F2810, TMS320F2811, TMS320F2812, TMS320C2810, TMS32C2811,
TMS320C2812”, Literature Number SPRS174S, Revised March 2011
[58] Texas Instruments, ”C28x IQmath Library, A Virtual Floating Point Engine”, Literature Number
SPRC990, Revised June 2010
104
[59] Beckhoff Automation, ”ET1100 Hardware Data Sheet”, URL
(http://www.beckhoff.de/german/download), Version 1.8, Revised 2010
[60] Rüedi H., Köhli P., ”Driver Solutions for High-voltage IGBTs”, PCIM Europe, Power Electronics
Magazine, April 2002
[61] Mihalachi M., Mutschler P., ”Position Acquisition for Long Primary Linear Drives with Passive
Vehicles”, IEEE annual meeting of the Industrial Applications Society, IAS 2008
[62] Khong P. C., Leidhold R., et al., ”Magnetic Guiding and Capacitive Sensing for a Passive Vehicle of
a Long-Primary Linear Motor”, 14th International Conference on Power Electronics and
Motion Control, EPE/PEMC 2010
[63] Rettenmaier T., ”Positionserfassung und Kommunikation zwischen zwei DSPs in modularen
Servoantriebssystemen”, Diplom Thesis Nr. 1349, Darmstadt University of Technology,
Institute of Power Electronics and Control of Drives, 2009
[64] Tomita M., Senjyu T., et al., ”New Sensorless Control for Brushless DC Motors using Disturbance
Observers and Adaptive Velocity Estimations”, IEEE Transactions on Industrial Electronics, Vol.
45, No. 2, April 1998, pp. 274-282
[65] Holtz J., ”Pulsewidth Modulation for Electronic Power Conversion”, Proceedings of the IEEE, Vol.
82, No 8, August 1994, pp. 1194-1214
[66] Seung-Gi J., Min-Ho P., ”The Analysis and Compensation of Dead-Time Effects in PWM inverters”,
IEEE Transactions on Industrial Electronics, Vol. 38, No. 2, April 1991, pp. 108-114
[67] Munoz A. R., Lipo T. A., ”On-line Dead-Time Compensation Technique for Open-Loop PWM-VSI
Drives”, IEEE Transactions on Power Electronics, Vol. 14, No. 4, July 1999, pp. 683-689
[68] Wang G. L., Xu D. G., ”A Novel Strategy of Dead-Time Compensation for PWM Voltage-Source
Inverter”, 23rd IEEE annual Conference and Exposition on Applied Power Electronics, APEC
2008
[69] Urasaki N. T., Senjyu T., ”Dead-Time Compensation Strategy for Permanent Magnet Synchronous
Motor Drive tacking Zero-Current Clamp and Parasitic Capacitance Effects into Account”, IEE
Proceedings of Electric Power Applications, Vol. 152, No. 4, July 2005, pp. 845-853
[70] Hyun-Soo K., Hyung-Tae M., et al., ”On-line Dead-Time Compensation Method using Disturbance
Observer”, IEEE Transactions on Power Electronics, Vol. 18, No. 6, November 2003, pp. 1336-
1345
[71] Leggate D., Kerkman R. J., ”Pulse-based Dead-Time Compensator for PWM Voltage Inverters”,
IEEE Transactions on Industrial Electronics, Vol. 42, No. 2, April 1997, pp. 191-197
[72] Leidhold R., Mutschler P., ”Speed Sensorless Control of a Long-Stator Linear Synchronous Motor
Arranged in Multiple Segments”, IEEE Transactions on Industrial Electronics, Vol. 54, No. 6,
December 2007, pp. 3246-3254
105
Curriculum vitae
Name Sorin Mihail Silaghiu
Birth date/place 31.08.1983/Romania
Nationality Romanian
Educational achievements
1997-2001 High School Liceul Nichita Stanescu, Ploiesti, Romania
2001-2006 Study of Electrical Engineering and Computer Science at Transilvania University, Brasov, Romania. Specialization: Automation Systems and Industrial Informatics
Professional experience
2006 Student trainee at Siemens (Power Generation), Erlangen, Germany
2007-2011 Assistant at the department of Power Electronics and Control of Electrical Drives, Darmstadt University of Technology, Germany
since January 2012 Development engineer at GIGATRONIK company in Stuttgart, Germany